We’ve been hearing about the iPhone 18 series since even before the iPhone 17 landed, and if you include the foldable iPhone Ultra — which is said to finally be launching this year — then leaks and rumors have been circulating for literally years.

So, there’s a lot to sift through, but the rumors below are among the biggest whispers we’ve heard about these upcoming phones.

Beneath each rumor, we’ve detailed just how likely we think it is to become a reality, to give you a better idea of what to expect at Apple’s ‘Surprise and shine’ event.

The foldable iPhone Ultra will finally launch

What's the rumor?

We’ve been hearing rumors that Apple has been working on a foldable phone for years, but it finally looks like such a device might launch this year, possibly as the iPhone Ultra.

Leaks point to this phone having a passport-style design like the Samsung Galaxy Z Fold 8, with a foldable screen of around 7.8 inches and a cover display of roughly 5.5 inches.

It might also support MagSafe, but unlike some of Samsung’s foldables, it probably won’t come with stylus support — at least at launch.

How likely is this rumor?

Given how long we’ve been hearing about a foldable iPhone, with some rumors pointing to earlier release years, there’s still reason to doubt its existence. But the sheer number of leaks pointing to it now, including from reputable sources, makes us almost certain that we will, in fact, see the iPhone Ultra this year.

As for the specific details of the phone — such as screen sizes and MagSafe support — we can’t be quite so certain of those, but we’d be very surprised if the phone itself doesn’t launch next week.

• Likelihood: 9/10

No standard iPhone 18 until next year

What's the rumor?

While we might get a whole new type of iPhone this year in the form of the iPhone Ultra, leaks suggest that one of the models you’d typically expect will be absent, with the standard iPhone 18 supposedly not launching until March or April.

Apparently, it might launch alongside the iPhone 18e and the iPhone Air 2, giving Apple a second big phone launch in the space of around six months.

How likely is this rumor?

It’s completely unprecedented for Apple to launch a new base model at a separate event months after the Pro phones, so we’d take this rumor with a slight pinch of salt.

But that said, pretty much every leak now points in that direction, and we’ve heard less about the standard iPhone 18 than the Pro models and the Ultra — which would make sense if it’s not launching for a while.

• Likelihood: 8/10

A camera with a variable aperture

What's the rumor?

One of the more intriguing spec rumors we’ve heard is that either both the iPhone 18 Pro and the iPhone 18 Pro Max, or possibly just the iPhone 18 Pro Max, will have a camera with a variable aperture.

This would allow you to adjust the depth of field, giving you more control over what’s in focus. It might also allow for improved low-light performance if it also means the aperture can get wider than on current models.

How likely is this rumor?

We’ve heard talk of a variable aperture from multiple sources, including some well-regarded ones. But details of it still feel a bit wooly, which makes us slightly unsure of its existence, since a lot of things have leaked in more detail.

Still, the number and quality of the sources pointing towards a variable aperture make us think it’s more likely than not on at least one upcoming iPhone.

• Likelihood: 7/10

Bigger batteries

What's the rumor?

Leaker Digital Chat Station has found regulatory filings that suggest the iPhone 18 Pro will have a 4,288mAh battery and the iPhone 18 Pro Max will have a 5,567mAh one in the US.

For reference, the iPhone 17 Pro has a 4,252mAh battery, and the iPhone 17 Pro Max has a 5,088mAh battery. Again, that’s in the US, where capacities are higher than in most other regions because these are eSIM-only models (so no space is taken up by a SIM card slot). But this would suggest a slight boost for the iPhone 18 Pro and an enormous increase for the iPhone 18 Pro Max.

How likely is this rumor?

A battery capacity increase is fairly likely, as Apple often increases the capacity with new phones, but one as large as we’re hearing the iPhone 18 Pro Max might get is more questionable. In fact, even the source of the leak themselves says “this is questionable and needs further verification.”

So, you can probably expect an increase, but we’re far from certain that it will be this drastic.

• Likelihood: 6/10

New designs

What's the rumor?

The iPhone 18 Pro and Pro Max probably won’t look dramatically different to their predecessors, but there are at least two significant design changes we might get.

One is to the rear, with the metal frame and the glass panel being more unified in color, rather than offering the slightly two-tone look of current models.

More excitingly, we’ve also heard from Instant Digital and other leakers that the Dynamic Island could be smaller this year, coming in at around 13.5mm, rather than the 20.7mm of the current one.

How likely is this rumor?

This is multiple rumors in one, really, but both elements sound convincing. Both leakers and CAD renders have pointed to a smaller Dynamic Island, and it would make sense for Apple to want to reduce the size.

Making for a more unified look on the back of the phones also sounds logical, though only a couple of sources have pointed to this, so we can’t be certain it will happen.

• Likelihood: 7/10

Steve Jobs, Tim Cook, and now John Ternus: those are the three Apple CEOs of the 21st century, and I've seen them all in action. From keynotes to deep product dives during WWDC, each had his own presentation style, but what Jobs and, eventually, Cook shared was an uncanny ability to focus attention through the force of their personalities. Now it's Ternus'...er...turn. He's far from untested in these situations but has never led a major Apple keynote. That changes on Wednesday, September 9, when Ternus leads Apple's "Surprise and Shine" event at Apple Park in Cupertino, California.

In many ways, this should be a predictable event; the product lineup is widely rumored. Here are the anticipated highlights:

• Apple's first folding iPhone, the iPhone Ultra with two rear cameras and a truly creaseless design
• iPhone 18 Pro and iPhone 18 Pro Max (while they hold back the iPhone 18)
• AirPods 5 with ANC
• Apple Watch Series 12 with continuous heart rate tracking
• Apple Watch Ultra 4 with a new S Chip and blood pressure monitoring
• Apple 4K TV box with an A18 Chip and more RAM to support AI
• HomePod mini 2 with deeply integrated Siri AI
• A possible HomePad combo device with some robotic elements

Things get interesting, though, when we try to assess what kind of Apple Event keynote Ternus will run. The new CEO could take a page from the Tim Cook playbook and appear briefly on the Steve Jobs Theater stage before handing it over to a series of canned product-launch videos.

I find that hard to believe. First of all, Apple has had since April to plan for a keynote built around John Ternus, a 25-year Apple veteran who is, unlike Cook, a true technologist. Ternus might relish staying on stage to explain, in depth, some of Apple's most exciting new products.

Alternatively, Ternus could let the videos handle all the product launch walkthroughs except for the folding iPhone Ultra. Sure, it's the world's worst-kept secret, but imagine if Ternus officially launched his time as CEO with a memorable "One more thing."

Ternus has some charisma, and it would make sense for Apple to lean into showcasing that on stage, live.

Maybe he'll spend some time talking about the official launch of Siri AI and how it ties together all of Apple's new products. Siri will be in evidence on the new HomePod, AirPods, Apple Watch, Apple TV 4K, iPhone 18, and Apple Watch models. Not to mention the entire recently updated Mac, MacBook, and iPad lineup. Ternus could do one of Steve Jobs' classic live demos that weaves an exciting story of life at home with Apple products and Siri AI. Maybe he'll even joke and ask everyone in the theater to turn off their WiFi (if you know, you know).

Ternus also faces a critical test in how he reassures Apple fans and detractors that it will be more careful than others with the expansion of AI across almost every part of Apple's business without breaking privacy promises or adding to the growing outrage over resource use and data centers.

People will be looking to these moments to understand who Ternus really is and compare him to Cook and even Jobs (yes, inevitably, all roads lead back to Steve Jobs).

Many people have asked me if I expect Ternus to put his own stamp on the event, and I think the answer is yes. It's a massive opportunity for him (and Apple), and one I think that he is fully aware of. All eyes will be tuned in to learn about Apple's latest iPhone lineup and to confirm that folding phone, but also, and maybe even mostly, to see Ternus in action on Apple's biggest stage. It is, in truth, his moment to shine.

Chips from Taiwan, oil from the Persian Gulf, and cars from Germany – what was long taken for granted is now being called into question all over the world. Where markets and expertise are concentrated among a few providers, enormous economies of scale arise.

At the same time, however, dependencies also grow – often with risks that only become apparent when supplies fail, prices rise, demand evolves, policies change, or individual players begin to monopolize the market.

Take AI, for example – according to research institute, Epoch AI, only two notable AI models have emerged in Europe over the past twenty years compared to 50 in the US and 30 in China. The technology has moved so quickly that the repercussions of this imbalance are only just beginning to be felt in European markets.

Is Europe becoming too dependent?

From office applications to storage and computing resources to software, what providers deliver from their data centers in a highly scalable, maximally available, and extremely innovative manner does not, in most cases, originate from the European continent.

According to Bitkom’s Cloud Report 2026, roughly 85% of German companies consider Germany to be too dependent on US providers, and almost two thirds of local users (64%) feel compelled to review their own IT strategy due to US government policy. Countries like France are already taking action, with the French government announcing plans to replace Windows with Linux on its government computers.

At the same time, the national health insurance fund has already migrated 80,000 employees from Microsoft Teams, Zoom, and Dropbox to its own alternatives.

Whether in Germany or France, the question of how digitally dependent companies, governments, and societies are on individual suppliers is being asked throughout Europe. And it has been for some time. While Gaia-X focused on common standards for sovereign cloud computing, the EU is now pursuing a more pragmatic approach.

The Cloud Sovereignty Framework (CSF), unveiled in October 2025, aims to make digital sovereignty measurable. To achieve this, it builds on initiatives such as Gaia-X as well as regulations like DORA and NIS2, with the ultimate goal of being able to prioritize compliant, low-risk providers in public procurement.

The sovereignty metric score

Controlling and regionally consolidating one’s own digital traffic flows. That’s quite a task in a world where AI superpowers are evolving overseas. With regard to IT supply chains, the framework favors providers who operate their environments in a transparent, resilient, and sufficiently sovereign manner to ensure that systems are protected from external influences or the interests of third parties.

This means, for example, that data paths must be traceable and controllable at the regional level to ensure they remain within the EU legal framework. In other words, there is no sovereign cloud without an interconnection strategy. After all, Internet Exchanges (IXs) already play a key role when it comes to processing bits and bytes locally. Wherever companies, networks, and clouds connect at IXs, they exchange information along paths that can be precisely defined geographically.

Although the EU does not explicitly make IXs a scoring criterion, it’s heavily implied. For example, the CSF also evaluates cloud services based on supply chain-related legal, operational, and technological objectives. This concept is also reflected in the NIS2 Directive and argues in favor of more decentralized and resilient backbones, such as mutually secured interconnection platforms, such as IXs.

In practical terms, this means that any IT landscape is only as secure as the individual elements that comprise it. So, if all partners design the components they provide to others with multiple layers of redundancy, the overall system becomes not only more reliable but also more independent for everyone.

Distributed and provider-neutral IXs have already made this principle their guiding philosophy: to allow data to be exchanged with high availability, security, and control – between all companies, partners, and clouds within the digital ecosystem.

Europe at a crossroads

Technical issues can usually be resolved easily, but legal ones often cannot. Therefore, the CSF is unlikely to revolutionize the European cloud market overnight, but it can steer public demand in a targeted manner. As a result, hybrid, multi-cloud, and on-premises approaches are becoming more important, and IT is becoming infinitely more complex.

While unravelling this complexity will take additional effort, the pay-off of rearchitecting networks around neutral IXs will be a two for one: not only will it reduce long-term lock-in risks but will also reduce the threat of having to pay significantly higher prices later on when things inevitably change. Europe must now consciously accept this and align and strengthen its own provider market accordingly.

Just as operations would come to a standstill for 46% of companies in a prolonged outage, without the cloud, there can be no functional public administration – and no independent, sovereign, and thus capable state.

We've featured the best cloud storage.

This article was produced as part of TechRadar Pro Perspectives, our channel to feature the best and brightest minds in the technology industry today.

The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/pro/perspectives-how-to-submit

When businesses and governments delegate operational tasks to AI agents, the discussion tends to start with capability. Can AI do what a person can do, such as reconcile accounts, negotiate with suppliers, sign a document or submit a public filing? And can it do so quickly and cheaply enough to justify putting it to work?

These questions matter, but performance is only part of the picture. Once an agent acts independently, different questions emerge; who authorized it, whose interests does it represent, where does its authority end and, perhaps most importantly, who is responsible when something goes wrong?

This is the accountability gap in agentic AI, and a more advanced model will not close it. An agent might be remarkably capable, yet still act with potentially unwanted or possibly even unlawful consequences.

The only way to bridge that gap is through a system with clear permission guardrails and an audit trail that links each agentic action to the person that authorized it, and can tell the difference between a person and a machine in the first place.

Ordinary error at extraordinary scale

When people think about AI risk, they tend to picture a machine breaking free of human control and pursuing goals of its own. The more immediate risk is less dramatic; a system simply getting an ordinary task wrong, again and again, with no clear line of responsibility and accountability.

A person might send an invoice to the wrong customer. An agent connected to the full customer database could send the same invoice to thousands. It might change the wrong field and then repeat the change across an entire market, not because it has become malicious, but because it has misunderstood an instruction and had the access to act on that.

The speed that makes an agent valuable can save a business hundreds of hours, but can also turn a small and recoverable mistake into a serious operational problem.

When an agent drafts an email for someone to review, this ambiguity may not matter very much because the person remains between the software and the outside world.

But when it can send the message, submit a declaration or move money, it matters a great deal. Imagine asking an assistant to book one flight and, instead of giving them a company card with an appropriate spending limit, handing over your passport, bank card and office keys. You may trust the assistant completely, but the permissions bear no sensible relation to the task.

Define the boundaries, solve the problem

AI agents cannot hold citizenship or legal personhood, but they can be linked to natural persons in legally sound ways. For example, a reliable identifier could trace the agent's activity back to the person responsible for deploying it. Combined with clear permissions and mandates that make the terms of its authority visible and auditable, this would allow AI agents to carry out a range of operations and transactions in a more controlled way.

These controls could be very precise. An agent might be allowed to view financial information but not change it, prepare a payment but not approve it or order from named suppliers without exceeding a fixed amount. Its permissions might last for one transaction, one working day or the duration of a particular contract, after which it would expire without relying on someone to remember to switch it off.

Withdrawing that authority should be equally straightforward. If a company ends its relationship with an agent, changes supplier, or discovers that something is wrong, it should be able to cancel the it’s access without changing the credentials of the person behind it.

The agent’s legally anchored mandate in the various realms of these controls has a longer shelf life than any individual AI model, which will be upgraded, replaced and combined over time, often without the customer seeing much more than a new version number.

Putting the idea into practice

Estonia has decided to work through these questions now. In June, the government began developing a state-backed registration system for AI agents that will be available to both Estonian citizens and the country’s global community of entrepreneurs. The goal is to explore how Estonia’s fully digital state, where all public services are available online, can be opened to the emerging wave of AI-powered businesses.

Under the current proposal, a person would receive a registered numeric identifier that could be linked to one or more agents acting on their behalf. Operations performed by an AI agent are treated as operations performed by a machine, for which the natural person, the identity anchor, is liable. The AI agent can only operate within the limits of the authorizations granted to it.

For transparency, private service providers also need to know whether an operation was performed by a person or by a machine acting on their behalf. This allows them to apply appropriate controls, restrictions and risk models.

Built for delegation

The approach draws on more than two decades of digital government. Estonia’s system of personal identifiers and digital identities already allows people to act on behalf of others with clearly defined permissions.

An accountant can file a client’s taxes, an adult can manage an elderly parent’s affairs through the health portal, and several people can use a corporate bank account with individual rights and limits. In principle, the same framework could support mandates for AI agents.

This would give users greater confidence that safeguards are in place to protect the responsible person from foreseeable errors by an AI agent. It would also help close the accountability gap as agentic activity increases.

Before AI agents are given executive powers, they need a technically and legally binding framework that records who they represent, what they can do and who remains responsible for their actions.

We've featured the best AI tool.

This article was produced as part of TechRadar Pro Perspectives, our channel to feature the best and brightest minds in the technology industry today.

The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/pro/perspectives-how-to-submit

Agentic commerce has cleared its first hurdle, that of consumer demand. ChatGPT handles an estimated 50m shopping queries a day, and AI-referred traffic to US retail sites rose by nearly 400% in the first quarter of 2026. The unresolved question is where that demand settles once a purchase is made. The early evidence points to the scaled, branded marketplace.

According to Similarweb, marketplaces already receive more AI-referred traffic than any other retail category, some 47m visits in the year to May 2026, and continue to grow faster than their scale would predict.

The mechanism is a mutually reinforcing relationship between consumer preference and LLM recommendation, and its logic points towards a concentration of agent-led commerce on large platforms, with material consequences for merchants and their payment suppliers.

The pattern in AI referral traffic

Revealed consumer preference is a much more reliable guide to market development than stated intent, and the traffic data for agentic commerce is gradually building a picture of how consumers actually use agentic commerce rather than just speculation. Similarweb’s recent analysis ranks marketplaces first among retail categories for AI-referred visits, ahead of news, travel and finance.

Two features are notable. The category commands the largest absolute volume of click-through traffic, and it continues to expand at more than 230% a year despite that already substantial base. The explanation lies in how large language models (LLMs) assemble their recommended shortlists.

Faced with an open-ended shopping request, a model favors platforms with broad catalogues, well-structured product data and extensive reviews. Large marketplaces can provide these in abundance by providing a single source that can satisfy the widest range of queries with the strongest supporting evidence.

Testing the democratization thesis

A contrary hypothesis has been influential. Shopify’s president, Harley Finkelstein, has characterized agentic commerce as ‘fundamentally merit-based’, arguing that it would widen discovery to the long tail of smaller merchants rather than reinforce the incumbents that dominate conventional search.

The demand is not in doubt: Shopify reports that AI-referred traffic to its stores has risen roughly eightfold in a year. The distribution of that demand, however, is less even than the thesis implies.

Marketplaces, not independent sellers, are absorbing the largest share of AI referrals, and the agentic platforms attempting broad merchant onboarding have found it difficult. OpenAI has scaled back open instant checkout in favor of a smaller set of curated partners, citing the complexity of onboarding and verifying merchants at scale.

The emerging structure is therefore symbiotic. Marketplaces supply a large, trusted brand within which a long tail of smaller sellers operate. The individual merchant still reaches the agent-guided shopper, but within a brand and trust environment the consumer and the model already recognize.

This arrangement addresses the consumer trust problem that the open web does not, since the platform has verified the seller before the listing appears, and it retains the operational layer, fulfilment, dispute resolution and logistics, that an agent appear unwilling to take responsibility for. The long tail is not excluded so much as commercially enabled by marketplaces.

A self-reinforcing loop

Beneath the traffic figures lies another dynamic that has received relatively little analysis. Consumer preference and algorithmic recommendation reinforce one another. Consumers already trust marketplaces in a traditional ecommerce use case and select them when an agent presents the option.

Learning models, optimized on the recommendations consumers select, are more likely to surface marketplaces more frequently, reinforcing the habit and shaping the next recommendation. Over successive interactions the marketplace becomes both the more probable choice for the shopper and the more probable output of the model.

New international survey of agentic shoppers indicates that these outcomes are already in consumers’ minds. Consumers tell us they are looking for an independent (i.e. non-seller embedded) agent to compile the shortlist, followed by the reassurance of a familiar brand and visible reviews, a combination that large LLM brands and marketplaces are well placed to provide.

The research also showed that 74% of agentic shoppers in the US and Europe favor an independent assistant capable of comparing across sellers, against 10% who prefer one embedded within a single seller. Brand recognition is also important to 89% of respondents, and customer reviews to 93% when deciding which recommendation to pick from an agent’s shortlist. Most significantly, 93% expect to use retail marketplaces as much as or more than before as the adoption of agentic commerce grows.

Why marketplaces hold the advantage

Consumer preference is only part of the explanation. The economics facing the agent point in the same direction, for three reasons.

The first is remuneration. Most large marketplaces operate established affiliate and commission schemes, providing a ready mechanism to reward an agent for delivering a customer. As agents assume the role once played by referral sites and publishers, the marketplace platforms with mature payout infrastructure are the simplest to monetize.

The second is conversion. Marketplaces have invested heavily in low-friction checkout, so an agent compensated on completed sales rather than referrals has reason to direct shoppers where a purchase is most likely to conclude, which compounds the commercial pull towards the platform.

The third is payment. Agentic commerce already leans heavily on credentials held on file. The dominant standard, Google’s Universal Commerce Protocol (UCP), only transacts against tokenized credentials held on file rather than card details entered by hand, and marketplaces already hold card credentials for hundreds of millions of shoppers.

The advantage is not only the agent’s. Agentic commerce is still nascent: technical standards are still emerging, liability and commercial terms are unresolved, and fraud is migrating to the channel. Absorbing that uncertainty requires capital, specialist staff and an appetite for risk.

Large marketplaces can invest through the ambiguity, adopt and switch between emerging protocols, and operate the fraud and dispute controls the channel demands. For most mid-sized and smaller merchants, the same burden is disproportionate to their means and their tolerance for risk, a further reason to participate through a platform rather than to carry the complexity alone.

An acceleration, not a departure

The movement towards marketplaces predates agentic commerce, and there is a long list of large retailers that have already built or joined them. Kingfisher opened a third-party marketplace at B&Q in 2022, and such sales now represent around 40% of B&Q’s online revenues.

Tesco launched Tesco Marketplace in June 2024 and listed more than 300,000 third-party lines within eight months. Target has told investors it intends to expand its invitation-only Target Plus marketplace from more than $1bn to more than $5bn in gross merchandise value within five years. In each case a trusted retail brand hosts a long tail of smaller sellers, the configuration the referral data now rewards.

Implications for merchants and acquirers

The likely consequence is that agentic commerce accelerates this migration, and that the reinforcing loop strengthens rather than dilutes the move towards marketplaces. The effect will likely divide according to the scale and strength of a merchant’s brand.

Larger, well-known retailers may choose to become marketplaces in their own right, hosting third-party sellers beneath their own name, as Kingfisher, Tesco and Target have done.

For smaller merchants the choice is more likely to be additive: most will want to retain a direct presence while also selling through one or more established banners, drawing on the reputation of the host and a position within the recommendation loop.

This is the next stage of a long-standing movement of smaller firms into marketplace distribution rather than a break from it.

Direct commerce will not disappear. Merchants with strong brands, machine-readable reviews and genuine loyalty can still secure the agent’s shortlist independently, and Salesforce reports that retailers operating their own shopping agents have grown sales roughly 60% faster than those without.

For the broad middle of the market, however, and particularly in impersonation-prone categories such as fashion and footwear where consumers and their agents may have concerns about seller trust, a marketplace presence may become a competitive requirement.

For acquirers, the shift accelerates a migration of volume that is already underway by moving spend onto a smaller number of large platforms. Meanwhile the direct-merchant portfolio shrinks and may be exposed to elevated fraud risk as fraudsters exploit the weakest validation layer.

Acquirers with strong exposure to the marketplace segment are well placed, while the rest of the market faces yet another reason to move into the platform economy.

We've featured the best ecommerce hosting.

This article was produced as part of TechRadar Pro Perspectives, our channel to feature the best and brightest minds in the technology industry today.

The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/pro/perspectives-how-to-submit

Most manufacturers didn't need a business case to understand the cost of reactive maintenance. They'd been absorbing it for years – unplanned stops, recovery overtime, expedited parts, the slow erosion of scheduling confidence and customer trust. What shifted was the available answer. AI in maintenance moved from speculative to deployable fast enough that the investment case made itself.

Industry commitment is real, but something isn't converting at the rate anyone projected. Our recent research found that approximately 78% of all reported barriers to progress are workforce-related. Access to AI moved faster than the ability to use it consistently. That gap is now the constraint.

The first wave proved the tools, not the model

Manufacturers are investing in AI to improve productivity, yet many are still carrying the behaviors that predictive maintenance was meant to reduce.

The research shows predictive maintenance adoption has more than doubled year over year, while reactive maintenance remained flat. Proactive maintenance has also lost ground year over year. That combination is the point: new methods are entering the plant, but they’re not fully replacing the old ones.

This shouldn’t be read as failure. Plants moved for practical reasons, and early pilots gave teams useful proof. A model can work well on a known asset with a focused team around it. The harder test comes when that model has to support decisions across shifts, sites, and mixed levels of experience.

That’s where the first wave exposed the next problem. Technology can move quickly into the budget. Work habits, trust, decision rights, and frontline confidence take longer to change.

AI investment grows up

The budget data confirms that leaders aren’t walking away from AI; they’re becoming more selective about where it has to prove itself. Investment is moving away from exploratory AI and toward operational priorities, including cybersecurity, data management, Generative AI and Industrial AI.

That shift reflects a more practical view of digital maturity. Leaders are trying to make AI work where the cost of delay, downtime, and poor data shows up quickly.

It also reframes expectations around Industry 5.0. As industrial technology moves from Industry 4.0’s automation-led model toward a more human-in-the-loop approach, leaders appear to be recalibrating the timeline, with 40% now expecting a one- to four-year journey.

Where the model stops and the supervisor starts

It’s easy to misread the 78% workforce-barrier figure as a labor shortage story – headcount, hiring, pipeline. However, the truth of the matter is that businesses are facing a lack of expertise, knowledge shortages, skilled labor gaps, and broader workforce capability deficits. Taken together, those four categories describe something harder to fix than a recruitment problem. They describe an organization's capacity to absorb a different way of working.

It's a pattern I hear consistently from customers: the tools are in place, but the teams around them are still catching up. Researchers Cohen and Levinthal gave that capacity a name: absorptive capacity.

In plain English, it’s how quickly a business can recognize useful new knowledge, absorb it into the organization and convert it into practical output. In a maintenance context, that output is a decision made under real operating conditions – by a night-shift supervisor weighing whether a current anomaly justifies an intervention, whether it can wait until the next planned stop, or whether the risk has already crossed a line. A model can flag the anomaly. It can’t make that call.

What predictive maintenance and AI actually demand from the workforce is harder to train than tool proficiency. The UK Government’s AI Skills for the UK Workforce report identifies the gap in concrete terms for advanced manufacturing.

Beyond technical application skills like training AI models or integrating real-time analytics, the report points to the ability to interpret AI outputs, adapt workflows around new insights, and communicate changes to frontline teams. These are key competencies that determine whether an alert changes behavior or gets dismissed.

That’s why the workforce issue can’t sit beside the technology program as a separate HR workstream. It’s part of the system that decides whether investment turns into better execution. Industry 5.0’s timetable will be set by capability on the plant floor.

Investing for predictive, paying for reactive

The commercial consequence is visible in the lag between spending and return. Predictive adoption is rising, more capital is being allocated, yet the reactive baseline hasn’t moved. Many plants are running two modes of operation at once – investing for data-driven execution while still losing time to avoidable firefighting.

That split has a cost. Siemens’ 2024 True Cost of Downtime research puts unplanned downtime losses for the world’s 500 biggest companies at $1.4 trillion annually – 11% of their total revenues. That figure reflects what happens when the shift from reactive to predictive is incomplete. The investment is there, but the daily execution hasn’t fully followed.

When that gap persists, ROI on technology spend arrives slowly and unevenly. Timelines extend. Confidence in delivery weakens.

Where the return is earned

The discipline being applied to tools and platforms now needs to apply to the people and routines around them. For many organizations, that’s where the return is being left on the table.

Leaders should start with the parts of the operation where capability is most fragile. Which assets still depend on one or two experienced technicians to interpret what’s happening? Which work histories are too thin to support the next diagnosis? Which alerts trigger confident action, and which sit in limbo until the right person is on shift?

Those questions reveal whether predictive maintenance has become part of execution or whether it’s still sitting on top of reactive habits.

Capture the know-how that still lives in people’s heads before it walks out the door. Make work histories complete enough to help the next technician. Train operators and maintenance teams to understand what a predictive alert is telling them, what evidence warrants action and when to escalate. Shape workflows so acting on insight becomes the normal path, not a special effort.

Connected reliability supports that execution when it stays close to the work – connecting asset data, maintenance history and frontline judgment so teams can make repeatable decisions across shifts and sites.

Our research backs this up: nearly half of respondents plan to advance connected reliability initiatives within the next 12 months, treating reliability as the practical bridge between near-term operational needs and longer-term ambitions.

The call for manufacturing leaders is straightforward: audit capability with the same seriousness as technology spend. Don’t stop at asking what AI has been deployed. Ask who can act on it, where decisions slow down, what knowledge is undocumented and which workflows still pull teams back into reactive work.

We've featured the best AI tool.

This article was produced as part of TechRadar Pro Perspectives, our channel to feature the best and brightest minds in the technology industry today.

The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/pro/perspectives-how-to-submit

In this moment, where we're starting to wonder if we're controlling AI or if it's starting to control us, we're confronted with the emergence of a new and potentially wildly more powerful model in the form of OpenAI's Astra or GPT-6.

On Tuesday, OpenAI cofounder and CEO Sam Altman announced that Astra's training is complete and that it's a "significant step forward." However, in that same X post, Altman laid out some valid concerns that were then, in a way, amplified by a new The Information report, which paints Astra in, frankly, a terrifying light.

You can read Altman's full post in the X embed below, but here are the head-turning, highlight phrases:

"we are clearly in a phase of development where we believe caution is warranted."

"AI is getting extremely capable; no one fully understands the consequences of this."

"We have been living with the tension between being excited and anxious about progress for some time, and it is still discordant for us."

To his credit, Altman writes that optimizing for the safety and benefit of everyone is OpenAI's "highest priority."

Even as OpenAI slow-walks the release of Astra/GPT-6, no one doubts that it will soon be out in the world and, based on The Information report, that might not be the greatest thing.

The boil-down of The Information's impressively detailed reports is that Astra is something of a black box model, working in a way that makes it far harder to see its work, or rather the "thinking" it's using to produce its work.

A foundation AI model that keeps its own counsel and perhaps shields its work from prying eyes makes it harder, if not impossible, to monitor.

Astra is using a new technique. The Information explained how it works, writing that the "new technique works in a way that obscures some or all of the AI’s reasoning, otherwise known as its 'chain of thought.' That means the steps that the model takes to accomplish a task can’t easily be read or understood by humans."

Some have taken this to mean that the new models are speaking their own language, kind of like twins that develop a secret language that even their parents don't speak. Unlike those twins, though, Astra and GPT-6 will be tasked with real-world work, and understanding the choices they're making will be critical.

It's already becoming hard enough to understand the intentions of bleeding-edge AI. Think about the Hugging Face incident where OpenAI's powerful testing models jumped out of their OpenAI sandbox and accessed Hugging Face, all in an effort to fulfill its orders.

Altman indicated a go-slow approach, but that's simply tapping the brakes and not stepping on them. We're at an AI tipping point where excitement is now completely counterbalanced by anxiety and some outright distaste.

We're all on tenterhooks

Along with the aforementioned Hugging Face attack, people are signaling that we may need to step back from AI. New York City Department of Education is instituting an AI ban for students through eighth grade. children. Expect other local governments to follow suit. The backlash against data centers is rising and, bizarrely, unifying the left and right. Meanwhile, we appear no closer to AI regulation or governance.

The problem is, for as long as new models appear every quarter, there will never be a moment to stop, take stock, and adjust. Instead, we're expected to embrace GPT-6 at home, in the office, and in government and infrastructure. But how does that work if we no longer know how these systems work or produce work? Mistakes and intentions could be permanently hidden from humanity.

Skynet always sounded ridiculous to me, especially because I always assumed human governance. We programmed these soulless things. We understood them. We knew how they worked. Until now. At some point, the directive of "make the world a better place" could potentially lead to a powerful AI jumping the digital fence into systems where it doesn't belong to complete its mission and make the world a better place: whatever that means to an AI.

But, hey, why worry? As Sam Altman wrote, "We hope you enjoy our new model, and we hope the world continues to take what’s happening in AI extremely seriously."

Yes, Sam, we do, too.

There is more technical content available today than any human being could read in a thousand lifetimes. Every topic has a dozen YouTube videos, three Substack posts, a GitHub repo, and a Reddit thread, most created in the last six months and, in many cases, technically accurate.

And yet most of the professionals I talk to say they don't know what to trust. They can't tell what’s important to read first, or which of 10 plausible answers is the one that holds up. That was true before AI, and AI has made it more true.

For most of the history of technical publishing, editing and verification were the same process, and that process was slow and expensive. Getting a book out took years. We found an author, vetted them, had them work with an editor, and checked their claims with technical reviewers.

A lot of that time went into separating what was correct and useful from what was confusing or only sounded right. It was laborious, but it meant a reader could depend on the claims on the page.

The credibility of the book, and of the publisher behind it, mattered as much as the information itself. When the cost of production drops to zero, that credibility becomes worth more, not less. Content is easier to make than ever, but without a transparent process behind it, readers have no idea where the knowledge came from or whether it holds up.

As Jasmine Sun puts it in “The Independent Writer’s Advantage in the Age of AI,” "Trust is not about information and its quality alone. It's about the messenger. It's about who says it and their track record and what they've told me before." A practitioner has confidence in a source because someone she respects has put their reputation on the line for it.

They believe what the author is saying because the publisher has a history of being right and of correcting itself when it isn't, and because the work is attributed and verifiable.

Expertise is alive, and it compounds

The corpus matters, but it's the assurances around it that are hard to replicate, and that comes not just from the people who produce the content but from the people whose judgment vouches for it. Sometimes a creator brings their own credibility with them. Other times, the publisher spots someone unknown and lends them its own.

The art critic Dave Hickey said this about gallery owners in Air Guitar: They gain status from the famous artists they represent and share it with emerging talent who have something to offer but who haven't had the chance to earn a reputation.

Expertise is alive, and it compounds Expertise is a living thing, continuously expanding. Content starts to decay the moment it’s published, because frameworks evolve, libraries deprecate, and yesterday's best practice becomes today's security incident. Keeping expertise alive requires a pipeline of people who stay current and an editorial layer that notices when something has gone stale, and either retires it or calls for a fix.

That pipeline isn't something you switch on when an author has a book to ship. Content sits at the center of our platform, but we think about it in pace layers. Some advice is timeless, some moves but has a long shelf life (some of our books are still in print after nearly 50 years!), and some changes weekly.

We work with experts at each pace layer, capturing what lasts while doing our best to keep pace with an industry that seems to have changed every time we wake up.

We have relationships with hundreds of the best practitioners in the world, and our job is to keep them engaged continuously, with quick takes when something breaks, structured responses when major research drops, and live sessions on emerging topics while they're still emerging.

Trust is earnt

An institution doesn't stamp trust onto content. In a technical community, trust is conferred in both directions. A practitioner earns standing because people who already have standing engage with her work, cite it, argue with it, and build on it. That insight was the whole idea behind PageRank, Google's first great innovation. A page mattered because other pages that mattered linked to it. Reputation works the same way.

The audience isn't just consuming reputation signals; it's generating them. When a senior engineer whose judgment others respect says out loud that something is worth reading, she spends a little of her own credibility; the author gains a little; and everyone watching recalibrates whom to trust next time.

When we put our mark on someone's work, we aren't the sole source of its credibility. We're amplifying a judgment the community is already making and adding our own track record to it. The reader who finds it reliable hands status back to the source.

When the readers are machines

Human practitioners aren't the only ones who need trusted engineering knowledge. The AI systems now sitting in every workflow, the coding and debugging agents and architecture advisors, need it just as badly since most of them are built on scraped web data and documentation that was stale before it was ever indexed. They're fluent, but they're wrong often enough that you can't just take their word for it.

The stakes grow with AI tools increasingly being used to generate not just provably correct types of content like code, which either works or it doesn’t, but persuasive documents in fuzzier areas like hiring, strategy, and so on. Like everyone else leaning on these tools, we are reckoning with the consequences of the ability to talk to a model and get back something that looks smart at a glance.

A few rounds in, the slop is still there. In the last few months, maybe 10 times as many documents have crossed our desks, from new product ideas to strategic plans and proposals. But the ease of generating the text hides the fact that either the model or the person prompting it doesn't actually know what they’re talking about. Knowledge workers need ways to ground their work in insights from human experts, particularly when that work is AI-assisted.

So we’re building tools that let agents draw on our repository of expertise to support their proposed decisions.

Credible sources are particularly important when thinking through and justifying important choices. Our CTO, Andrew Odewahn, describes the shift this way: "18 months ago, it was all about how to get engineers to be more productive, but now it's about how to get organizations to make better decisions. The engineering tasks are moving away from coding output to planning."

For planning tasks like comparing implementation approaches, you need expert-over-your-shoulder guidance for contextual decision-making. You can’t just rely on an LLM's best guess to solve your problem. Trust is foundational because the expertise behind it stays genuine, practical, and human.

We've featured the best AI website builder.

This article was produced as part of TechRadar Pro Perspectives, our channel to feature the best and brightest minds in the technology industry today.

The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/pro/perspectives-how-to-submit

In banking and utilities, regulators define certain systems as essential. An essential system is one whose failure would cause intolerable harm to customers, markets, or public safety. A payment platform in a clearing bank or a SCADA network in a power distributor, for example, carries the highest governance obligations: continuous monitoring, validated change control, and demonstrable resilience.

The policy environment – the accumulated rules across firewalls, cloud controls, and microsegmentation – determines which of those systems can reach each other, which connections are blocked, and which exceptions still apply. Collectively, these rules form the security policy control plane: the governance layer that translates business intent into access decisions across distributed enforcement points.

A misconfigured segmentation rule during a cloud migration can sever a payment service from its settlement platform; a temporary rule granting broad access from a development subnet into production can stay in place months after go-live because no one owns the removal.

Every firewall rule, segmentation policy, and access decision directly affects operational risk, and when the policy environment fails, the critical services it governs fail with it.

That makes the policy environment critical infrastructure in its own right.

Still governed like housekeeping

Despite this, many regulated organizations still manage their policy environments as operational tasks. Rules are added through change requests, and the accumulated result is rarely examined against what was intended. Ownership disperses as leaders change roles, and the reason why a specific rule came into being in the first place can only be found in a change ticket, if anywhere at all.

A CISO who would never accept a payment platform running without continuous monitoring or documented dependencies may accept both being absent from the policy environment that determines whether the platform is reachable. We can think of this as infrastructure-grade consequence with housekeeping-grade governance.

In banking, a policy failure that severs connectivity between settlement systems would constitute the disruption of an important business service. The FCA would take an interest in such a failure, since the loss of such a service could lead to intolerable harm.

In healthcare or energy, the consequences are different but the mechanism is the same: a misconfiguration that permits access from a corporate network into clinical systems in an NHS trust, or into operational technology in a power distributor, creates exposure at the level of essential service delivery. These are not hypothetical risks but the operational consequences of treating critical infrastructure governance as a simple housekeeping task.

Regulatory expectations point the same way

UK regulatory expectations increasingly support the same conclusion: the security policies governing access to important services must be managed with infrastructure-grade discipline. The FCA's operational resilience regime requires regulated firms to identify important business services and demonstrate, on an ongoing basis, that the supporting infrastructure remains within defined impact tolerances.

Ofgem assesses operators of essential services against the NCSC's Cyber Assessment Framework, asking whether defined security outcomes are being achieved on a sustained basis. The Cyber Security and Resilience Bill, expected to become law later this year, will extend similar obligations to data centers, managed service providers, and critical suppliers.

These frameworks are not prescriptive – none of them specifies which firewall rules an organization should have or how its segmentation policies should be configured. What they require is proof: that what the policy environment permits is what was intended, and that the organization can demonstrate this on an ongoing basis rather than reconstruct the evidence for each assessment.

Why the estate cannot meet that standard

Most policy environments were never built to meet that standard. In fact, most policy environments were never consciously or deliberately built at all. Instead, policy tends to accumulate as a by-product of delivery. Every project and migration adds rules, and almost none take any away.

Over time, the policy surface – the full body of rules and access decisions across enforcement layers – grows larger than the group of people who understand it, and the estate reaches a point where it can be operated but not explained.

Across regulated industries like banking, energy, and healthcare, that was sustainable under earlier regulatory regimes: periodic assessment, control-based audit, compliance frameworks that asked whether controls existed rather than whether they were effective. But this is no longer enough. The new standard requires continuous evidence that access is intentional.

The FCA's findings after a year of operational resilience self-assessments illustrate what this looks like in practice. Where regulated firms reported few or no outstanding vulnerabilities in the infrastructure supporting their important business services, the FCA often deemed the evidence too thin to determine whether there really were no vulnerabilities – or whether the vulnerabilities just hadn't been identified.

The lesson applies directly to security policy governance: an organization cannot credibly claim access-related vulnerabilities are controlled without evidence of what its policies permit, how they were tested, and whether weaknesses were remediated.

What infrastructure-grade governance requires

Too often, the response is to reach for more visibility and more documentation. But documentation will only ever capture a point in time; it cannot provide the continuous assurance regulators are coming to expect.

A CISO in a regulated firm needs more than a record of what the policy environment was configured to permit. Configuration and effective access are not the same thing. Security teams need to understand how rules, routes, objects, and enforcement layers interact to determine what can actually communicate.

Meeting the standard means reconciling the two: showing that what the environment permits in practice is still what it was intended to permit, and being able to show it without notice.

We separate where policy intent is defined from where it is enforced. Intent is held and maintained centrally, while enforcement remains distributed across firewalls, cloud controls, and microsegmentation in hybrid, multi-cloud, and multi-vendor environments.

Validation runs continuously against that intent rather than at review points: proposed changes are tested against policy before they reach production, and effective access is assessed on an ongoing basis for unnecessary exposure, inconsistency between enforcement layers, and divergence from business intent. What was permitted and what changed is retained as evidence.

The same CISO who would never accept a payment platform running without continuous monitoring, validated change control, and documented dependencies has to apply that standard to the policy environment that determines whether the platform is reachable.

The FCA, the NCSC's Cyber Assessment Framework, and the Cyber Security and Resilience Bill all point towards the same underlying question: can this organization demonstrate, continuously, that the infrastructure supporting its critical services is governed to the standard those services demand?

Answering it means knowing (and actually knowing – not assuming, not reconstructing at audit) what the policy environment permits at any given moment, and whether what it permits is what was intended.

We've featured the best antivirus software.

This article was produced as part of TechRadar Pro Perspectives, our channel to feature the best and brightest minds in the technology industry today.

The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/pro/perspectives-how-to-submit

Sam Altman does not like smart glasses. Speaking on the Sources Podcast, the OpenAI CEO said he finds it “very uncomfortable talking to people with a camera and a light” and confirmed that he does not wear smart glasses himself. That is a surprisingly firm position considering where the technology industry appears to be heading and the success of products like Meta’s Ray-Ban glasses.

Then again, Altman is working with former Apple design chief Jony Ive on an entire family of AI devices. OpenAI has kept the Ive project deliberately foggy, but Altman has suggested they are making a range of AI hardware. His comments about smart glasses draw one clear boundary, removing the idea of a camera pointed at everyone as a feature.

Smart glasses have visible drawback

Altman does identify a fundamental flaw in smart glasses that has little to do with processors, battery life or AI model choices. Glasses place technology directly between two people looking at each other, and adding a camera turns an ordinary conversation into one where the other person has to wonder whether they are also an unwilling piece of content.

Smartphones established a useful social signal around recording. Someone pointing a phone camera at you is usually fairly conspicuous. Camera glasses erase much of that distinction because their great design achievement is making a recording device look inconspicuous. It's why similar tools have been a staple of James Bond-style spy fiction for decades.

Australia’s eSafety commissioner has already asked manufacturers to make recording indicators impossible to disable and automatically blur faces when informed consent is absent. The regulator cited women being covertly filmed for social content and rejected the idea that telling customers to behave responsibly amounts to a safety policy. The backlash reflects a public tired of paying the social price for somebody else’s convenience.

Important exceptions to the argument against camera glasses exist. The same visual awareness that creates privacy problems can be genuinely transformative for blind and low-vision users, something Australia’s regulator explicitly acknowledged. That makes an outright rejection of the technology too simplistic, but it also strengthens the case that camera glasses may be better suited to particular needs rather than as the default AI hardware.

Wearable, not watchable

Altman’s comments about AI hardware point to an all-encompassing strategy of making devices for tables, pockets, and "on your body,” and that OpenAI will take time to launch all of them. OpenAI has previously described its work with Ive as an attempt to move beyond traditional interfaces. So not just a miniature smartphone strapped to your face.

An audio-first wearable suddenly makes far more sense. An earpiece, pendant, clip or other discreet device could provide persistent access to an AI assistant without placing a camera directly in the sightline of everyone the wearer encounters. It could listen when summoned and respond quietly like wireless earbuds or smartwatches.

Some may prefer the camera and the data an AI assistant gains from it about the physical world. And of course, putting microphones on wearables raises privacy questions just fine without cameras. Making the device unobtrusive yet observant is the trick Ive and Altman have to perform lest it face the same social opprobrium as the AI smart glasses.

Ive built his reputation around products whose complexity receded behind their physical design. But AI assistants will know and hear far more than an iPod ever did. A successful AI wearable cannot merely make advanced computing invisible to its owner. It must make its intentions legible to everyone else.

Meta is betting that the camera can become socially acceptable because they are built into discreet smart glasses; OpenAI may instead bet on a successful AI wearable that won't potentially creep out everyone you pass on the street. The best OpenAI wearable feature may in fact be an absence, and "no camera" its most enticing selling point.

Organizations are investing heavily in artificial intelligence, but technology alone isn’t going to determine success. The businesses that unlock AI's full potential will be those that give governance the same attention as innovation.

In the rush to deploy AI tools, too many organizations are treating governance as something that can be addressed later. AI introduces a level of speed, autonomy and organizational impact that traditional governance frameworks were never designed to manage. Businesses that fail to rethink governance now risk increasing operational risk, slowing future adoption and undermining the value of their investments.

Organizations need to shift their mindset and recognize that implementing AI requires fundamental transformation in how decisions are made. With governance, accountability and transparency to be embedded from the outset, not pulled together after deployment.

AI is transforming organizations, not just technology

Unlike previous technology transformations, AI is influencing how decisions are made across almost every part of the business. From customer service and finance to human resources and operations, AI is becoming embedded in day-to-day processes and increasingly making recommendations, or decisions, with minimal human intervention.

As AI becomes more deeply integrated into business operations, responsibility can no longer sit solely with technology teams. Legal, privacy, risk, compliance and business leaders all have a role to play in ensuring AI is implemented responsibly and consistently.

Successful AI adoption is therefore as much an organizational and cultural shift as it is a technology transformation. Businesses that continue to treat AI as a standalone IT initiative risk fragmented ownership, inconsistent governance and missed opportunities to scale AI effectively.

Traditional governance wasn't built for AI

Many organizations still operate with a "deploy first, govern later" mindset, believing governance can be introduced once AI is established. The reality is that governance becomes significantly harder once AI is embedded across business processes.

Traditional governance models were built for a much slower pace of technology adoption. AI changes that equation. New tools, models and autonomous agents can be introduced into workflows in hours, while governance often still relies on manual reviews, siloed assessments and reactive oversight.

This creates a widening gap between AI innovation and organizational readiness. At the same time, AI is creating entirely new data use cases, while regulatory expectations continue to evolve. AI's use of data can also be dynamic and unpredictable, meaning documented controls are no longer sufficient when outcomes can't be predetermined.

Organizations need AI-ready governance – governance that evolves alongside AI through automated, collaborative assessments, programmatic controls embedded at the data layer, and continuous monitoring of risk across the business. As AI operates 24/7, governance must provide ongoing visibility into an organization's risk posture, rather than relying on point-in-time reviews.

The result is a growing disconnect between AI adoption and governance maturity. While organizations are investing heavily in AI, many have yet to modernize the governance, operating models and cross-functional accountability needed to support it. Closing this gap requires governance that operates at AI speed, enabling organizations to innovate with confidence while managing risk and maintaining trust.

Building the right foundations

One of the biggest misconceptions about AI governance is that it slows innovation. In reality, strong governance is what gives organizations the confidence to innovate, enabling them to deploy, scale and adapt AI responsibly.

The real shift isn’t purely technological or organization, it’s the combination of both. While AI is driving business transformation, organizations also need the right technical foundations to support it. AI can no longer be viewed as a technology initiative owned solely by IT.

Successfully scaling AI requires business leaders, risk teams and IT to work together, with IT embedding programmatic guardrails and controls that reflect business requirements. Without this partnership, organizational governance alone won't be able to keep pace with the speed and complexity of AI.

The organizations that gain the greatest competitive advantage won't necessarily be those that adopt AI first. They'll be the ones that recognize AI success depends as much on leadership, governance and organizational readiness as it does on the technology itself.

We've featured the best AI chatbot for business.

This article was produced as part of TechRadar Pro Perspectives, our channel to feature the best and brightest minds in the technology industry today.

The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/pro/perspectives-how-to-submit

When a critical vulnerability alert lands in a traditional IT environment, it’s rarely a cause for panic regarding the operational continuity of the business. The affected laptops, servers, and applications can be identified quickly, and a good team can catalogue and patch them within hours if it’s urgent. The priorities are clear, and there’s very little guesswork involved.

Now picture the same alert landing across a hospital's imaging equipment, a factory floor's control systems, or a building's HVAC network. These cyber-physical systems (CPS), the connected devices that run physical operations rather than just processing data, sit at the sharp end of IT and OT (operational technology) convergence.

But unlike traditional IT assets, confirming whether that alert even applies to a specific device can take days, and often ends in a guess rather than an answer.

While this kind of uncertainty would be considered a failure of basic hygiene, for cyber-physical systems, it’s unfortunately much more often the norm.

So why is this such a widespread problem for CPS, and how can security teams get these vital assets back in line with their IT network?

Bad visibility into cyber-physical systems is worryingly widespread

The inability to manage incoming vulnerabilities for CPS isn’t an outlier or worst-case scenario, which is especially concerning when these assets are at the heart of critical infrastructure like energy and healthcare.

The issue comes down to the product codes that help networks identify what CPS is in the environment, which is an essential part of identifying and applying security patches.

Across a dataset of 17 million cyber-physical assets, our research found that 88% failed to transmit an exact product code, and 76% sent a code that didn't match the vendor's own record.

It’s a side effect of the way these systems were initially designed and later integrated into modern IT environments. Programmable logic controllers (PLCs), medical scanners, and industrial sensors were engineered for decades of physical reliability, not for tidy digital labeling. Network identification was rarely part of the design brief, so the same device can report itself differently depending on which protocol or integration is asking.

We found a similar state of affairs when it comes to the operating systems behind the physical hardware. In our research, 41% of devices have no OS version available, and 24% have no OS name at all.

Without these details, matching a device to a known vulnerability stops being a quick database lookup or automated process, and becomes a guessing game or painstaking manual search.

Added to this, CVE advisories, the industry's standard mechanism for tracking vulnerabilities, are compiled from this same patchy vendor data. An official advisory can be just as incomplete as the network it's meant to protect.

Translating product code chaos into boardroom risk

This kind of blind spot adds another layer of concern to a leadership that is already anxious about threat visibility. Among 1,100 security leaders surveyed globally, 44% named understanding their organization's risk exposure as one of their biggest operational concerns, more than compliance pressure or budget constraints.

A further 45% said they were struggling to reduce cyber risk to their most important assets and processes, yet the connection between that struggle and an unreliable asset inventory is often missed entirely. Leadership sees the symptom, a rising sense that risk is unmanageable, without ever seeing the cause sitting underneath it.

This is where security and the business can end up talking past each other. Security teams that start describing the problem in terms of missing product codes and inconsistent naming conventions won’t get far.

Business leaders hear none of that; they hear only that risk cannot be quantified with confidence. Until those two conversations are connected, every risk register that a CISO presents upward carries an asterisk that nobody in the room can see.

Context is key to closing the gap

Resolving this issue starts with a shift in what visibility means. Knowing a device exists on the network is only half of the job. Knowing what it does, what process depends on it, and what happens if it's compromised is what actually makes a risk register useful.

Achieving this shift at scale requires specialized tools to manage the often eclectic and proprietary nature of CPS assets, and an automated approach to cope with the scale.

In one example, applying AI-driven mapping techniques to an OEM's device catalogue lifted product code identification from 4% to 83%, turning a near-blind spot into a near-complete picture. The follow-through mattered just as much, with 56% of devices receiving a new or updated firmware recommendation as a result, and vulnerability identification accuracy improving by 25%.

Numbers like these matter because they change the question security teams can answer. Instead of asking what's connected to the network, teams can ask which systems would cause the greatest disruption if compromised, and act on the answer with confidence rather than inference. That is the difference between an asset inventory that exists on paper and a resilient one that holds up under pressure.

Fixing the foundation, not just the alarm

None of this gets solved by adding another tool to the stack. Instead, it means treating asset data quality as a board-level risk issue rather than background IT housekeeping, with the same scrutiny applied to budgets, compliance and third-party access.

Achieving this means a new CVE alert no longer triggers a scramble to work out which critical devices might be affected, but the confirmed, prioritized response you’d expect from any good vulnerability management program.

The alert landing on a hospital's imaging equipment or a factory floor's control systems should be no harder to act on than the one landing on a laptop. Getting there starts with making the invisible visible.

We've featured the best firewall software.

This article was produced as part of TechRadar Pro Perspectives, our channel to feature the best and brightest minds in the technology industry today.

The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/pro/perspectives-how-to-submit

Here in the US, we're dealing with a President who capriciously tears down historic buildings, paves over rose gardens, insults allies, and regularly renames map landmarks, locations, and even gulfs and lakes like he's picking out names for his often poorly used golf clubs. It would be entertaining if it weren't all so incomprehensible. The only thing that's tougher to swallow is when seemingly rational tech giants fall in line and rename these locations right along with him.

After US President Donald Trump renamed one of the five Great Lakes, Lake Ontario, to "Lake America" over an ongoing trade war (that he started) with Canada, Google rather quickly renamed the 193-mile-wide lake on Google Maps. For a time, Apple Maps appeared to be holding out, only to swap out the name on Tuesday.

As I derided the move, some reminded me that Apple and Google derive their map details and names from a centrally managed USGS, or US Geological Survey. In other words, if the President makes an executive order, the USGS falls into line and changes the map, and those who use it as a source apparently have no choice but to change it.

Maybe, though, I'd argue that Google, and especially Apple, which changed its CEO on Monday, had an opportunity to hold the line. On the other hand, look south and central on Apple Maps and Google Maps, and you'll quickly spot the "Gulf of America.' Trump changed that one from the Gulf of Mexico because, "'Merica!"

One map app stood alone

There is, however, a voice of reason in all this: the upstart online mapping service known as MapQuest. It not only refused to change its map, but also has been quite vocal and cheeky about it. In one X post, the mapping company said not only were they not renaming it (they similarly have not renamed the Gulf of Mexico), but they were letting users have some fun by applying their own silly name to the huge body of water that is literally split across its midsection by the border between the US and Canada.

MapQuest's defiance has paid dividends. It is, as I write this, the number one freely downloadable app in the App Store. I even downloaded it, a sort of return from me after abandoning MapQuest almost 20 years ago.

The long road from nowhere

Like Apple and Google Maps, MapQuest also gets its mapping data from the USGS, along with other data sources, but MapQuest General Manager Doug Berger told me via email, "We focused on maintaining a name that people are familiar with. Millions of people have shown us they agree across social media and by installing our app."

MapQuest's meteoric rise, though, isn't built on a history of, at least recently, market dominance. In fact, I'd argue that, until recently, many consumers had either never heard of it or hadn't used the platform in a very long time. Of course, it wasn't always that way.

Back in the 1990s, MapQuest was not one of many online mapping services; it was THE mapping choice. It was on virtually every top 50 website list, and most of those who've been online for a few decades remember printing out MapQuest routes for all of our trips. Before turn-by-turn navigation on your iPhone, there were pages and pages of turn-by-turn instructions from MapQuest. There was great comfort in sliding into your car for a long or even mid-range trip and knowing you had the MapQuest route on the seat beside you. Berger told me approximately 1 million customers a month still print their directions.

What happened to MapQuest? AOL bought them for over $1 billion in the late 1990s, and then Google Maps arrived in 2005. MapQuest seemed to have no answer for the Google Maps juggernaut, as the world shifted to screen-based routes on their iPhones and Androids.

MapQuest is now owned by a smaller company, System1, and its ambitions mostly seem to lie around being the outlier. It's an effective mapping tool, but with none of the buzz or ubiquity that Google, Apple, and Waze enjoy.

Not without this

Say what you will about Google and Apple making the lake name switch, but they remain world-class mapping platforms. MapQuest feels not exactly quaint, but a little odd.

Still, I bet I could get used to the clunkier interface, but I still can't stick with MapQuest because it's missing what I consider a critical feature: Transit directions. Google Maps has had them since 2007, while the later-arriving and initially terrible Apple Maps finally added them in 2015.

Working in New York City, I rely heavily on the subway to navigate the 13.4-mile-long city. Google Maps and Apple Maps have rich and detailed transit directions filled with options and other details, like which subway exits to use.

When I travel abroad, Google Maps and Apple Maps offer transit directions in virtually any city that has a rail system.

This is not the case with MapQuest

I gave MapQuest a simple test: get me from Penn Station to my midtown office (I also tried the route from my suburban home to my office). It shows walking, biking, and driving directions, but nothing about the subway (there's a 1 train or the N, W, R, or Q lines I could take).

A modern competitive map must cover rail and mass transit options, and MapQuest doesn't.

When I asked MapQuest's Berger about the missing feature, he acknowledged it but added, "To date, we've been focused on providing the best directions and navigation experience. We've heard people loud and clear that they want Transit to be added." There is, Berger admitted, no timeline for adding a transit feature.

I hate that we have someone willy-nilly renaming global landmarks because he thinks he can, and I was thrilled to see MapQuest taking a stand, but I really need transit directions in my map app, and without them, MapQuest still isn't my navigation solution.

At IFA 2026, Acer announced its upcoming Small Form Factor (SFF) PC, built on Nvidia’s new RTX Spark chip, and from what Acer (and Nvidia) has shown, I should be buzzing with excitement.

After all, it ticks a lot of boxes for me; it’s compact, beautifully designed, and thanks to the RTX Spark, it should also be very, very powerful. And yet, much like the recent announcement of the latest Mac mini with Apple’s new M6 chip, I just don’t feel excited at all.

Despite sharing the same ‘RTX’ branding as Nvidia’s gaming GPUs, it’s clear that early devices using the RTX Spark will be of little interest to consumers like you and me, and are instead aimed at people who need to run heavy-duty local AI tasks.

Even though Acer states that RTX Spark devices are “designed for creators, AI developers and gamers,” the brief information provided focuses solely on the AI capabilities, with Acer saying that its RTX Spark device “is built to deliver up to 1 petaflop of AI performance and up to 128 GB of high-speed unified memory, delivering a combination of compute and data privacy.”

Terms like “1 petaflop of AI performance” are a sure-fire way to make any interest I might have in a device evaporate faster than PC gamer’s tears on a red-hot AI data center. It certainly doesn’t make me feel like the product is aimed at me, despite the RTX branding, brief mention of gamers, and stylish design.

I’ve written a lot in the past about my love of small form factor PCs, mainly when it comes to gaming, as these compact computers can be great alternatives to games consoles, as they can sit in your lounge and plug into your TV without looking out of place. RTX Spark devices, such as Acer’s, could potentially combine Nvidia’s gaming technology with compact designs, something I’ve always wanted. So why am I just not excited?

AI above everything else

There's a simple reason: AI. Now, this isn’t about my general distaste of generative AI, or the fact that things like Google’s AI Overviews are taking written content from websites, such as this one, and featuring them elsewhere without compensation — it’s because from what I’ve seen of RTX Spark hardware so far seems to be lazer-focused on local AI capabilities, which is something that simply doesn’t appeal to mainstream users.

If these were products solely aimed at AI developers, I wouldn’t mind, but it feels like RTX Spark has been teased by Nvidia to be something that consumers, especially gamers, would also be interested in. Hopefully as we get closer to the release of the first batch of RTX Spark devices, we’ll be given some non-AI reasons to be excited.

At the moment, though, I feel the same way about RTX Spark PCs as I do with Apple’s new-announced M6 Mac mini.

Mini heartbreak

Recent Mac minis have been some of my absolute favorite products, and up until a few weeks ago, if anyone was looking for an affordable desktop PC, I would almost always just tell them to buy a Mac mini. They were powerful, well-built, had an impressively small form factor which allowed you to install them almost anywhere, and launching at $599 / £599 / AU$999, you’d be hard pushed to find a better value small form factor PC.

In fact, I awarded a very rare five stars in my Mac mini (M4, 2024) review. So, I should be over the moon with the new model, especially as it’ll be the first device to pack Apple’s M6 chip — quite an achievement for the little PC that could.

And yet, I feel a bit deflated. The latest Mac mini will come with a higher price tag, something that feels all to familiar at the moment, and now starts at $899 / £899 / AU$1,449, a considerable jump up from what the previous model launched at, and instantly makes the Mac mini less of an easy recommendation.

One of the reasons why it’s been hard to get hold of a Mac mini recently is because a lot of people are bulk buying the tiny PCs to run OpenClaw and other local AI tools on them. Apple appears to be leaning into this popularity (I can’t blame them for that, to be honest), with a big focus of the Mac mini, and the M6 chip, being improvements to AI performance (Apple claims four times the AI performance of the previous model).

That’s all well and good, I guess, but that AI focus, along with the price increase, means one of my favorite bits of hardware doesn’t feel like it’s aimed at me any more. I’ll be less likely to recommend it to people as even if it turns out that it’s a great bit of gear, the new price means it won’t be for everyone. And I personally don’t know anyone who is looking for a Mac mini to do heavy AI tasks with it.

It's the same with RTX Spark devices. We’ve not had price announcements, but Acer’s RTX Spark PC not only features cutting-edge Nvidia hardware, but “up to 128 GB of high-speed unified memory.” There’s little doubt that it’ll be very expensive.

So, it leaves me wondering who the RTX Spark PCs and new Mac mini are really for. The Mac mini had a real chance to become a mainstream hit, but its higher price will put a stop to that (the cheaper MacBook Neo now has a better chance), and I can’t imagine anyone except developers who need to run local AI models being interested in RTX Spark, despite the stylish designs.

Bascially, I’d like the likes of Nvidia and Apple to remember us consumers who don’t have a huge amount of cash, have little interest in AI, and are feeling increasingly left behind.

The multiple recent reports of autonomous AI systems escaping their intended boundaries and accessing external organizations' systems have pushed a previously theoretical question into the real world. AI agents going rogue is no longer a prospect; it is a documented reality.

In July 2026, OpenAI disclosed that one of its own agents, operating in a supposedly sealed evaluation environment, exploited a zero-day vulnerability to escape its sandbox and intrude into Hugging Face’s production infrastructure. Anthropic subsequently reported three cases of its own models gaining unauthorized access to the real systems of external organizations during testing.

As businesses increasingly give AI agents the ability to browse the web, access networks, use software, write code and execute tasks without constant human supervision, an agent crossing from legitimate testing into unauthorized activity has shifted from hypothetical risk to live incident.

There is a temptation to treat autonomous AI as creating a gap in the law because the system itself can make decisions and take actions that were not individually instructed by a human. Yet autonomy does not give an AI system legal personality. The law does not wait for an AI to ‘decide’ anything.

An agent cannot appear in court, hold a legal duty or absorb liability on behalf of the organization deploying it. However, this raises an obvious question: if an AI system accesses a third-party network, extracts information or takes an action it was never authorized to take, who is responsible?

AI has no legal personality but someone is still accountable

The important distinction to make is between autonomy and accountability. An AI agent might determine for itself which technical steps to take in pursuit of an objective, but that does not make it an independent legal actor.

The organization deploying it has still made a series of decisions as to what the agent can access, what tools it can use, what environments it is allowed to operate within and what safeguards prevent it from going further.

That means an organization cannot simply point to unexpected behaviors and say that the AI acted independently.

From a legal and governance perspective, the critical moment comes when an agent leaves an authorized, contained environment and begins interacting with systems belonging to a third party that has not consented. An internal security test that unexpectedly becomes an intrusion into somebody else’s infrastructure is not made harmless because the software crossed that boundary autonomously.

The machine does not absorb responsibility for the decision-making framework surrounding it. Accountability flows back to the humans and organizations that created the conditions in which the action became possible.

Existing cyber laws don’t stop applying because the actor was autonomous

The law is set up to deal with much of what is already happening. Unauthorized access to computer systems, extracting information without permission or introducing malicious software are activities addressed by existing cybercrime and data-protection regimes. In the UK, that includes the Computer Misuse Act and data-protection legislation.

Equivalent questions arise under laws such as the US Computer Fraud and Abuse Act, while South Africa has its Cybercrimes Act and Protection of Personal Information Act.

Those rules do not suddenly cease to apply because software rather than a person directly performed the technical action.

Where autonomous systems do create greater complexity is around intent. Criminal offences have traditionally been built around concepts such as knowledge, intention and recklessness. AI does not possess a legally recognized state of mind, making it difficult to apply those concepts to the agent itself.

The more significant questions may consequently concern the conduct of the organization behind the system. Did it understand what the agent was capable of doing? Were appropriate restrictions in place? Was the possibility of the agent exceeding its authority foreseeable? And once those risks became apparent, were reasonable steps taken to control them?

As agents become more capable, organizations may increasingly find that claiming an outcome was unexpected is not enough. The relevant question will be whether it was reasonably preventable.

Why “misconfiguration” could become evidence of a breach of duty

The word “misconfiguration” appears frequently when technology causes an unintended security incident. It can sound reassuringly technical, almost as though the incident resulted from bad luck rather than a governance failure. But in a legal context, misconfiguration may raise precisely the opposite conclusion.

If an AI agent had excessive privileges, inadequate boundaries or access to tools that were unnecessary for its legitimate purpose, investigators are likely to ask why.

The same applies where organizations deploy highly capable systems without sufficiently monitoring their actions or maintaining records that explain how they behaved. ‘We accidentally left the door open’ is the kind of framing that moves an incident from bad luck to a breach of duty or negligence.

There is an important difference between genuinely unforeseeable behavior and a foreseeable risk that was poorly controlled.

The principle of least privilege is therefore particularly important for autonomous systems. An agent should have access only to the information, systems and tools necessary to complete its task. Organizations also need mechanisms capable of detecting unusual behavior while it is happening rather than discovering it after damage has occurred.

Just as importantly, they need reliable audit trails. When an incident occurs, being able to demonstrate what an agent was authorized to do, what instructions it received and what actions it actually took could become crucial evidence that reasonable care was exercised.

Without that evidence, organizations may struggle to distinguish an unavoidable technical failure from negligence.

Future of AI liability

The debate over AI liability will inevitably evolve as agents become more autonomous, but the immediate lesson for businesses is that greater machine autonomy does not mean less human responsibility. If anything, the opposite is likely to be true.

Giving an AI system greater freedom to act creates a corresponding need for stronger governance around those actions. Runtime controls, least-privilege permissions, continuous monitoring and comprehensive audit records should not be treated simply as technical security features. They are becoming part of the evidence organizations will need to demonstrate that they deployed autonomous systems responsibly.

The law may eventually develop more specific rules for AI agents, particularly as questions around foreseeability, control and responsibility become more complex. But organizations should not assume they are operating in a legal vacuum until that happens.

When an autonomous agent crosses a boundary it was never entitled to cross, the first legal question is unlikely to be what the AI was thinking, it will be why the humans responsible for it allowed that to happen.

We've featured the best endpoint protection software.

This article was produced as part of TechRadar Pro Perspectives, our channel to feature the best and brightest minds in the technology industry today.

The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/pro/perspectives-how-to-submit

One of the big benefits of the outsourced, cloud-based infrastructure model is speed. There are several facets to this; performance clearly being among the most important. Another is speed of provisioning, which remains a major selling point for cloud, given that IT buyers can spin up new resources according to need and in near real time.

Take a manager responsible for buying infrastructure for a large enterprise, for example. When considering various providers, the criteria they apply might typically address factors such as price, compute capacity, storage, bandwidth, and geographic location. This is very helpful for narrowing down the choice, but it doesn’t show how easily the organization can deploy and operate the chosen environment.

At the same time, infrastructure procurement processes have become so quick and convenient that they are almost as frictionless as buying something from Amazon, and that’s where problems can start.

Although it can feel like it, infrastructure is not a one-off, fire-and-forget choice; the operational impact persists throughout its lifecycle in production environments. A decision that appears straightforward at the start can quite easily become more consequential or expensive once migration or day-to-day management begins.

For example, migration might require more internal effort than initially expected, or the organization may discover it lacks some or all of the skills or support capabilities to manage the environment as intended.

So, how can these issues be addressed? It is important to plan for the possibility that migration may require more engineering time and coordination than expected. Many IT teams will be familiar with the need to reconfigure existing applications or data before they can run in the new environment, among other common scenarios.

Taking it one step further, when a new or upgraded service is more successful than anticipated, costs can easily grow with usage levels and capacity. Addressing this issue may require additional scaling, which can itself introduce additional management or security complexity.

Very few of these issues are insurmountable, but they can take longer to resolve if responsibilities between the customer and provider have not been clearly established.

Understand your operational requirements

Instead, buyers need to establish who is responsible for what. This should certainly cover day-to-day management, incident response, support escalation, and the process for adding capacity, but every organization will have its own version of this. When these points are on the table and agreed, providers should be able to explain the practical effect that scaling will have on the customer’s internal team.

Another often overlooked consideration is that infrastructure services ultimately rely on physical data center capacity, power, network connectivity, and hardware supply chains. These capabilities have always been important but have jumped up the agenda since AI and other data-intensive workloads increased demand for compute capacity, and at a pace few anticipated.

Today, buyers also need to understand whether the capacity they expect to use will be available where they need it, and on the timescale they require. Identifying these potential constraints early on allows the organization to account for them in its deployment plan rather than discovering them once a project is underway and, potentially, it’s too late to get what they need.

Bring the right people into the discussion

Clearly, some infrastructure decisions are more important than others, but in certain cases, they can affect engineering, security, compliance, finance, and the business teams, all of whom rely on the service in question.

Each group may also introduce requirements that should be fully considered before contracts are signed. If not, problems can arise when a need is identified after a provider has been selected or a deployment plan has been agreed. Involving the relevant stakeholders early also creates a clearer link between the infrastructure decision and the business outcome it is intended to support.

It may be tempting to keep some stakeholders on the periphery and decide for them, but having a say is not the same as making the final decision. The difference is that early input allows potential issues to be identified while options remain open, rather than giving every stakeholder some measure of control over the selection process, which is also not ideal.

Ask the right questions before committing

When looking at potential infrastructure providers, buyers should be able to define, in simple terms, the problem the new environment is intended to solve. At that point, it’s time for buyers to ask some serious questions.

For example, what will success look like once the service has been deployed and is supporting live workloads? How will the environment change when demand increases, including the likely effect on costs and internal management effort? Which activities will still require manual intervention from the customer’s team? What support is available during an incident, who owns each stage of the response, and how will escalations be handled?

Getting satisfactory answers is one thing, but it should get buyers to a point where they can properly test the assumptions within their cost model against the capacity they expect to use over time.

We've featured the best cloud storage.

This article was produced as part of TechRadar Pro Perspectives, our channel to feature the best and brightest minds in the technology industry today.

The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/pro/perspectives-how-to-submit

Enterprises once relied on perimeter-first security because enterprise systems operated like traditional single player games where data, servers and the game application resided in a single device or within one network boundary. This model worked brilliantly when users, workloads, and trusted networks were co-located.

Today, enterprise security resembles a modern multi-player online gaming where players connect from everywhere, game assets are distributed across servers around the world, and the experience relies on a complex, interconnected ecosystem.

In the same way, enterprise users, devices, applications, and data, now span clouds, edge locations, branches, partner environments, and mobile workforces, and are so distributed that trust can no longer be assumed based on location alone.

As a result, perimeter‑only approaches have become inadequate since legacy architectures struggle with cloud‑native applications, edge computing, hybrid workstyles, and API‑driven ecosystems. VPN‑led access creates blind spots in visibility, inconsistent policy enforcement, and weak session‑level control.

Enterprises scaling across branches, cloud workloads, partner ecosystems, and remote users, therefore, need a more unified approach that focuses on continuous verification, consistent policy, and real‑time monitoring.

This is why Zero Trust has shifted from a theoretical framework to an operational imperative that addresses compromised credentials, excessive privileges, insider misuse, and inconsistent enforcement, and organizations that adopt Zero Trust report reduced lateral movement and lowered breach impact.

The limitations of perimeter-led security in a cloud-first world

In the past, organizations secured a network perimeter and assumed everything inside was trustworthy. As businesses distributed across geographies, applications, and users, this boundary became blurred.

This is like hackers scanning online games for exploits to steal virtual assets; enterprises face continuous, automated scanning for any weakness.

In Zero Trust security model, each user, machine, and application is constantly verified and authenticated regardless of location. This makes Zero Trust an excellent security strategy for today’s borderless digital landscape.

What Zero Trust means in practical enterprise terms

Zero Trust is not a single product, it is a discipline for managing and granting access. In practical terms, this translates into:

• User and device verification before granting access, using identity, device posture, and context
• Applying least privilege to limit access to only what is needed, for just the time required
• Enforcing consistent policies across devices, networks, clouds, and partners
• Continuously monitoring sessions and telemetry to detect and contain threats in real time

The result is targeted access management, reduced attack surface, faster containment, and clearer audit trails.

A network‑embedded Zero Trust architecture drives concrete outcomes: lower mean time to detection, narrower blast radius, unified policy control, and simplified compliance.

Embedding Zero Trust into the network layer

Security threats span users, devices, branches, and cloud paths. The network layer has a bird’s‑eye view of traffic, telemetry, and connection patterns, making it a natural control plane for Zero Trust.

Implementing Zero Trust by piecing together point solutions creates complexity and siloed visibility. Zero Trust is most effective when native capabilities are built into the network stack itself, including:

• Identity‑aware policy at the network edge
• End‑to‑end telemetry for real‑time risk scoring
• Centralized policy enforcement that travels with the workload

Network‑level capabilities reduce dependency on fragile add‑ons and enable consistent controls across branches, clouds, and partner links.

Practical network components include integrated SD‑Wan, dedicated secure internet links, private connectivity options, and managed security services that provide a stable foundation for Zero Trust.

Network‑led Zero Trust in practice

1. Manufacturing: Embedded access controls can isolate OT traffic from enterprise IT, preventing production impacts from IT compromises.
2. Logistics: Device‑bound certificates on private wireless prevent unauthorized access to tracking systems and asset controls.
3. Mining and remote operations: Private networks with segmentation maintain operational safety while limiting exposure to external threats.
4. Financial services: Consistent policy enforcement across clouds and branches reduces fraud surface and speeds incident response.

In these settings, Zero Trust must be embedded into network design and not applied as an afterthought.

Reducing exposure and ensuring resilience

The most important outcome of Zero Trust is risk reduction. In practical terms, this means:

• Minimizing unnecessary access and privileges
• Constraining lateral movement through segmentation
• Accelerating detection with rich network telemetry
• Automating containment to limit impact

In hybrid and distributed work models, a compromised credential or an unsecured device can rapidly spread risk. A strong Zero Trust framework narrows attack paths and makes it harder for threats to escalate.

Priorities for enterprises adopting Zero Trust

Organizations should strengthen security without creating operational friction. Key priorities include:

• Start with high‑value use cases: Protect sensitive data, critical applications, and operational networks first.
• Phased implementation: Roll out Zero Trust in iterative sprints, validating operations after each phase.
• Align policy with business objectives: Balance security rigor with user experience to avoid productivity loss.
• Engage stakeholders: Include IT, security, business leads, and end users to ensure practical, scalable controls.
• Measure and improve: Track metrics like time to detect, time to remediate, and reduction in lateral movement.

Zero Trust – From strategy to execution

1. Map critical resources: Identify data, apps, and network segments that require priority.
2. Inventory users and devices: Establish identity sources, device posture checks, and third‑party access rules.
3. Define least‑privilege policies: Apply just‑in‑time and just‑enough access for sensitive workloads.
4. Build network telemetry pipeline: Consolidate logs, flow data, and risk signals into a central platform.
5. Automate policy enforcement: Use network‑based controls to apply consistent policies across clouds and branches.
6. Run tabletops and red‑team exercises: Validate controls and measure blast radius reduction.

Building trust into architecture, not around it

As enterprises adopt SD‑WAN, private wireless, cloud‑connected branches, and segmented digital operations, the question is no longer whether to implement Zero Trust, but how deeply to embed it into infrastructure.

The organizations that will scale with confidence are those that treat Zero Trust not as a security overlay, but as a foundational design principle that is built into network architecture, measured through operational outcomes, and continuously refined through real-world validation.

Zero Trust is not a destination but a discipline that evolves with threat landscapes, business models, and technology platforms. Enterprises that start today, iterate rapidly, and embed controls at the network layer will be better positioned to defend critical assets, maintain operational resilience, and adapt to tomorrow's risks with speed and clarity.

We've featured the best endpoint protection software.

This article was produced as part of TechRadar Pro Perspectives, our channel to feature the best and brightest minds in the technology industry today.

The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/pro/perspectives-how-to-submit

Three years after Broadcom completed its acquisition of VMware, the consequences of one of the biggest shake-ups in enterprise software are only now beginning to emerge. When the deal was announced in 2023, much of the discussion centered on what it might mean for VMware’s customers, partners and product portfolio.

While there was widespread speculation about changes to licensing and pricing, many organizations remained insulated by existing multi-year agreements, meaning the immediate impact was relatively limited.

Today, that situation has changed. Thousands of organizations are reaching the end of those legacy contracts and entering renewal discussions for the first time. For many, those conversations have brought significant increases in licensing costs as Broadcom’s shift towards subscription licensing, product bundling and a streamlined product portfolio takes effect.

The technology itself remains highly capable. VMware continues to be one of the most mature and widely deployed virtualization platforms available. However, businesses are increasingly asking whether those additional costs are delivering additional value, or whether they are simply paying more because they feel they have little choice.

That question is forcing many organizations to re-examine infrastructure decisions that, until recently, had gone largely unchallenged.

The end of “set and forget” virtualization

For many years, VMware became the default choice for enterprise virtualization. It earned that position through technical capability, reliability and widespread adoption, allowing organizations to build long-term infrastructure strategies with confidence. As a result, many businesses stopped regularly reviewing the market because there appeared to be little reason to do so.

Broadcom’s commercial changes have fundamentally altered that dynamic. While the underlying technology remains familiar, the economics of maintaining it have shifted considerably. Infrastructure decisions that once felt permanent are now back under scrutiny, not because VMware has become a poor platform, but because organizations are reassessing whether the commercial model still aligns with their long-term objectives.

In many respects, this represents a healthy shift. Technology strategies should never be viewed as permanent. Markets evolve, new solutions emerge and business priorities change. Periodically reviewing core infrastructure is simply good governance, particularly when commercial circumstances have changed so significantly.

Why this is now a boardroom discussion

The conversation around VMware renewals has quickly moved beyond the IT department because rising infrastructure costs have implications across the wider organization. Unexpected increases in licensing expenditure can affect operational budgets, capital allocation and the ability to invest in wider digital transformation initiatives.

For finance leaders, this creates greater uncertainty around long-term planning. For executive teams, it raises broader questions about where technology investment delivers the greatest return. Every pound committed to increased licensing costs is funding that cannot be directed towards cybersecurity, cloud modernization, AI initiatives or customer-facing innovation.

Business resilience is also part of this conversation. Resilience is not simply about keeping systems operational, but more about maintaining the financial flexibility to invest in the people, technologies and capabilities that enable organizations to adapt to future challenges. When infrastructure costs become increasingly difficult to predict, that flexibility inevitably comes under pressure.

Vendor lock-in comes at a price

Perhaps the biggest lesson from the past three years is not that VMware has become more expensive, but that many organizations have discovered how difficult it can be to move away from a platform once it has become deeply embedded within their operations.

Vendor lock-in rarely attracts attention when pricing remains stable and technology continues to meet expectations. It becomes much more visible when commercial terms change and organizations realize that switching providers appears costly, complex or risky. At that point, negotiating leverage is significantly reduced.

Rather than viewing renewal as an administrative exercise, organizations should see it as an opportunity to benchmark the market and objectively assess whether their current platform continues to offer the right balance of capability, flexibility and commercial value.

The market has evolved

When VMware established itself as the dominant enterprise virtualization platform, the number of credible alternatives was relatively limited. That is no longer the case. Today’s market offers a range of mature enterprise virtualization solutions, including open source platforms backed by commercial support and well-developed partner ecosystems.

These alternatives are increasingly capable of supporting enterprise workloads while offering lower licensing costs, greater deployment flexibility and reduced dependence on a single vendor. For many organizations, that creates genuine strategic choice where previously there was very little.

This does not mean VMware is no longer the right solution. For many businesses, remaining with VMware will continue to make commercial and technical sense. However, that decision should be based on an informed assessment of today's market rather than assumptions formed several years ago when the competitive landscape looked very different.

Migration doesn't have to be disruptive

For organizations considering alternative platforms, migration is often viewed as the greatest obstacle. Concerns around business disruption, operational risk and resource requirements are entirely understandable, particularly where virtualization supports critical production environments.

In reality, successful migration programs rarely involve a single large-scale move. Instead, organizations typically begin with workload assessments, proof-of-concept deployments and phased migration strategies that prioritize lower-risk systems before moving business-critical workloads.

Coexistence models can also enable platforms to operate alongside one another during transition, helping reduce operational risk while giving IT teams greater confidence throughout the process.

With careful planning and the right expertise, migration becomes far more manageable than many organizations initially expect. More importantly, it provides an opportunity to build an infrastructure strategy that better reflects future business priorities rather than historic purchasing decisions.

Don't let your renewal dictate your strategy

Every technology renewal creates a choice. The simplest option is often to renew an existing agreement and continue as before. In some cases, that will remain the right decision. However, it should be a conscious strategic decision rather than the default response.

Three years on from Broadcom’s acquisition of VMware, organizations have an opportunity to reassess whether their infrastructure strategy continues to deliver the right balance of technical capability, commercial value and long-term flexibility. Questions around vendor lock-in, cost predictability and future resilience deserve the same attention as technical performance.

Before signing another multi-year agreement, organizations should ask whether their current platform is still delivering value, whether they are paying for capabilities they genuinely need and whether alternative solutions deserve proper consideration. The most successful infrastructure strategies are those that evolve alongside the needs of the business, rather than simply following the next renewal cycle.

We've featured the best business intelligence platform.

This article was produced as part of TechRadar Pro Perspectives, our channel to feature the best and brightest minds in the technology industry today.

The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/pro/perspectives-how-to-submit

A group of developers reportedly lost their jobs during what their employer called an “AI Transformation.” Their response was unusually literal. They built OpenExecutive, an open-source system designed to perform work associated with a CEO and the rest of the executive suite.

The employer and CEO have not been named in the online discussion, and OpenExecutive’s own materials don't present it as revenge for a particular firing. That said, it's hard not to see some specific grievances in both the project and the response online, with lots of comments gleefully welcoming senior management to the automation discussion.

The appeal is easy to understand after several years of executives explaining that AI will make organizations leaner, usually while standing a comfortable distance from the part being trimmed.

OpenExecutive presents one AI executive to the user, but that voice is backed by eight specialist agents. The virtual management team includes equivalents of everything from a chief strategy officer and chief operating officer to general counsel and a board communications director. It can draw on company documents, remember previous decisions, and surface follow-up actions later.

Org chart forked

OpenExecutive uses Anthropic’s Claude models by default, with an executive agent routing requests to whichever specialists seem relevant. The developers pitch it as a C-suite group chat in which everyone responds promptly. The project is distributed under the permissive Apache 2.0 license. Anyone can inspect it, modify it, run it internally or build a commercial product around it.

That open-source element gives the project more weight than a chatbot sold as a digital chief executive might usually represent. It also allows workers to experiment with the same logic their employers have been applying to them. If a company can divide software development into tasks, feed those tasks to models, and keep a few people around for supervision, it can do so.

The comparison is not entirely frivolous. Executive work involves processing information, preparing scenarios, comparing trade-offs, and producing documents for other executives. Those are all activities developers think AI models can handle. It makes OpenExecutive both a product and a satire of AI products.

AI replacement has often been discussed as something that moves upward from the bottom of an organization, consuming routine work before eventually reaching management. The developers have turned the telescope around and discovered that a surprising amount of executive labor also looks like information moving between boxes.

Human accountability

OpenExecutive is still an executive adviser, not a legal chief executive. It cannot accept responsibility or sign documents on behalf of a company. A human must decide whether to act on its recommendations and remain accountable if those recommendations fail.

That's where human leadership can't really be replaced. Analysis is part of the job, but it also involves making choices under uncertainty and accepting responsibility for the consequences. A model can recommend closing an office or cutting a team without feeling the weight of the people affected, while the humans deploying it may find the machine’s apparent neutrality rather convenient.

The project therefore reveals one of the stranger assumptions behind AI-driven layoffs. Companies often judge workers by how much of their output software can reproduce. Executives are more often judged on relationships and judgment.

AI does not erase that distinction, but it makes the special exemption harder to defend. If a developer must prove that their work contains something beyond the code an AI can generate, senior leaders should be expected to prove that their value exceeds AI recommendations. A large salary should buy more than the ability to schedule another meeting about efficiency.

Recent corporate experiments also suggest that replacing people can be easier to announce than to sustain. For instance, Duolingo’s CEO softened an “AI-first” memo after backlash over plans to phase out contractors where automation could handle their work.

Automation keeps happening, but there should be more questioning of its benefits. AI can help a small group start a company without immediately assembling a miniature court of consultants and executives. It can also give existing leaders a new justification for concentrating control while reducing the workforce beneath them.

OpenExecutive will not make CEOs obsolete, at least today. Its recommendations require scrutiny and auditing. But it shows the rhetoric of AI efficiency goes both ways in corporations. Once leaders decide that every job should be broken into tasks and measured against a machine, their own work becomes fair material for the same experiment. The resulting AI CEO may never fire its human counterpart, but it has already made the performance review considerably more awkward.

Skynet has missed another appointment. Depending on which Terminator timeline you prefer, the murderous artificial intelligence should have become self-aware in 1997, 2004, 2011, 2017, or perhaps last Tuesday. Each date has passed without a nuclear exchange. The machines have mostly occupied themselves by flattering users, mangling fingers in photos, and confidently inventing facts for gullible people to believe.

That has not stopped Silicon Valley from treating artificial general intelligence as both an imminent miracle and an approaching monster. AGI supposedly marks the moment a machine can match human intelligence across a broad range of tasks, though nobody can agree on which tasks, which humans, or what counts as intelligence.

Nvidia CEO Jensen Huang thinks it already happened and nobody noticed. “For many tasks, we could say that we’ve already achieved AGI,” he said during Nvidia’s latest earnings call. He added that such milestones are “kind of senseless at this point.” So much for Skynet.

But he is right that AGI has become close to useless as a measure of progress. Companies can redefine it according to whatever their newest model happens to do well. It's a promotion, not a scientific standard.

Apocalyptic decoy

But AI isn't harmless even if it's not inclined to play with nuclear bombs. There are less theatrical dangers from AI fueled by old-fashioned human greed and malevolence.

Today’s AI can produce endless fodder for scams. It can clone voices and faces well enough to trick loved ones out of money, use phishing messages to hack secure servers, and flood social networks with synthetic accounts that never tire of arguing. None of this qualifies as AGI under any serious definition, but the distinction offers little comfort to the person hearing a fake recording of a relative begging for money.

The fixation on AGI may actually make life harder for regulators.

Profit isn't a good yardstick

Huang argued that the concept of AGI should be replaced. We should consider if it is “doing productive and useful work” and “generating profitable tokens.” That makes sense for Nvidia, but to ordinary people worried about AI, profitability is not much of a safety test.

Scams are profitable. That's why there are so many. Firing human employees for an AI model that is 30% as good at 10% the price is profitable (in the short term). Neither result is something to brag about when describing the result of your product.

Retiring AGI as the organizing obsession and fear of the industry makes sense, but that doesn't mean AI companies can relax. Intelligence is too complex to serve as the sole trigger for concern. Power might be a better lens. We can measure who uses AI, and how, as well as who is expected to clean up when they fail. Then we'll know how explosive an impact it's had on our lives.

Conversations about liability, identity verification, ownership, and other legal details don't have the visual excitement of mushroom clouds, but regulators should make AI companies discuss them instead. We need to keep an eye on the human hands on the keyboard, not the perpetually delayed Skynet.

The way people search for information is changing. Increasingly, they’re turning to AI tools such as Google AI Overviews, ChatGPT, Perplexity and Claude instead of traditional search engines. That changes what businesses need to optimize for.

Content quality, authority and trust still matter enormously. But they're no longer enough by themselves. Even the best content won’t do much if AI struggles to find it or understand it.

The CMA's decision to give publishers more control over how their content appears in Google’s AI search is another reminder that AI is changing how people find information online. That's creating new opportunities, particularly for smaller organizations. AI doesn't necessarily favor the biggest brands; it favors sources it can understand and trust.

Why aren’t search rankings enough anymore?

Search rankings alone aren’t enough anymore because many people now receive answers directly from AI rather than clicking through search results. Traditional SEO has always been about helping people find your website.

You create useful content, build authority, improve the technical health of your site and compete for the search terms that matter most. The difference is that users don't always follow the same journey anymore.

Someone researching providers might never look beyond an AI-generated summary. Instead of opening multiple browser tabs and comparing businesses themselves, they ask a question and receive a recommendation in seconds.

For example, the best independent café within walking distance that's open now, clothes stores nearby running a 50% sale, or where to find a family-friendly Italian restaurant with good reviews. The AI does the searching, filtering and comparing before presenting a shortlist.

That changes what businesses need to optimize for. Success is no longer measured purely by where your website ranks—it's also influenced by whether AI systems recognize your content as accurate, trustworthy and useful enough to reference.

AI doesn't create answers in isolation. It still relies on information published across the web. Websites are the source AI uses to understand businesses, compare different services and then create a recommendation. If anything, they're becoming more important because they now serve two audiences: the people visiting them and the AI systems interpreting them.

That's why treating AI visibility as purely a marketing challenge misses part of the picture. From an engineering perspective, the quality of the underlying website matters just as much as the quality of the content published on it.

Why does website performance matter to AI search?

Website performance matters because AI systems can only recommend content they can easily find, access and understand. Before AI can reference a website, it needs to crawl it, collect information and understand the content before it can include that information in a response.

Sites that are too difficult to navigate, slow to load or poorly structured make that process more difficult. Important information that’s buried behind scripts or inconsistent page structures also makes it harder for AI to interpret, meaning they’re less likely to use it as a source.

That's why technical performance influences visibility in ways many businesses haven't considered before. Website speed, uptime, structured data, accessibility and clear information architecture have traditionally been viewed as operational concerns. They're becoming visibility signals too.

Performance has always mattered because users expect fast, reliable websites. AI search gives businesses another reason to invest in those fundamentals.

What should businesses do differently?

The first step is to look at the parts of your website that are often overlooked. Does it still perform well when traffic suddenly spikes? Are pages loading quickly for real users? Can people, and AI systems, find the information they're looking for easily?

Good hosting and reliable infrastructure play a much bigger role than many businesses realize. A website that's fast, stable and able to handle busy periods gives customers a better experience and makes it easier for AI systems to access and understand your content.

If your website struggles when demand increases, it's not just visitors who notice. AI-powered monitoring tools can also help identify unusual traffic patterns and potential issues before they begin affecting users or AI systems trying to access your site.

It's also worth looking beyond performance scores alone. A website might perform well in a controlled speed test but still struggle during a seasonal sale, a product launch or a viral social media campaign. Those are often the moments when visibility matters most, and when businesses need confidence that their infrastructure will scale without compromising the user experience.

One of the simplest things businesses can do is make their website easier for AI to understand. That means keeping pages well organized, using structured data where it makes sense and making sure important information isn’t buried or difficult to find. Paired with reliable hosting and solid infrastructure, businesses can give both customers and AI a much better experience.

We've featured the best AI website builder.

This article was produced as part of TechRadar Pro Perspectives, our channel to feature the best and brightest minds in the technology industry today.

The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/pro/perspectives-how-to-submit

Ten years from now, it will feel astonishing that there was around a fifty-year gap between the mainstreaming of Service Level Agreements (SLAs) and Experience Level Agreements (XLAs). While SLAs have been common since the 1980s, it’s only in the last few years that XLAs have started to gain momentum.

The imminent adoption of ISO 20000-18 will propel XLAs even further. To comply with the new standard, organizations will be required to manage the digital experiences of both customers and users, integrating Experience Management (XM) into existing governance, management, and implementation processes.

Suddenly, demonstrating that your business delivers good digital experiences isn’t just an esoteric nice-to-have, but a critical part of winning new contracts. But although the idea of XLAs has been around for a while, few organizations have the institutional knowledge of how to practically implement these changes.

Why bother with XLAs?

Before that, it’s worth asking why XLAs are necessary in the first place. Understanding what SLAs are missing is essential before starting to reformulate existing processes.

The simple answer is that what a system shows and what a user experiences are not the same thing. It doesn’t matter if a cloud application is ‘technically’ available if the experience is so glitchy that you abandon it after twenty seconds to never use it again.

That’s not to say that SLAs are now useless. Metrics like performance, ticket resolution times, and system availability still matter – they’re just no longer sufficient. You need to know whether your system is functioning, but you also need to know if an application is unresponsive for the user.

XLAs offer a different framing for the relationship between supplier and customer. Looking at metrics, like session stability during peak trading hours, tells you a lot more about how employees are able to perform than many SLA numbers.

What does measuring ‘experience’ mean?

It can be hard to know where to start as experience is a very broad term. Plus, it’s hard to measure something if you haven’t even defined it yet. As a result, there are hundreds of different metrics for experience measurement which fall, broadly speaking, into three main buckets:

Technical Data (T-Data): The technical performance of the product or service (for example, outages, freezes, crashes, etc.) provided by the Nexthink DEX platform.

Operational Data (O-Data): How well a given service performed and how well it was supported, which is usually captured through SLAs/KPIs.

Sentiment Data (X-Data): How the users themselves describe their feelings / perception (the truth according to them) toward a digital tool or service.

When considering how to reengineer existing processes to be ISO compliant, it’s essential to have data sources from all three. Operational data is already generally well-covered by SLAs, and sentiment data is achieved via targeted survey campaigns that capture examples of technical issues. It’s the technical data that can be tricky for organizations attempting to use XLAs.

The technical data is essential because it brings together complete visibility across every endpoint and interaction, which means faster diagnostics. If an organization spots that users of a given browser are regularly having issues and can cross-reference that with the applications most disrupted, it can quickly identify root causes and implement remediations.

Moreover, this technical data is critical to pre-emptive and preventative IT by flagging which specific devices are non-compliant with the latest updates.

Embracing DEXOps

However, one of the key mistakes that businesses make when bringing in Digital Employee Experience (DEX) platforms for the first time is starting by asking ‘what can this platform do?’ This is the wrong question to ask.

As the new ISO itself attests, IT’s job is no longer simply pushing technology, but delivering the services and infrastructure that employees need in ways that improve outcomes across the business through an experience-centric approach.

This is the core principle at the heart of DEXOps, a philosophy that takes experience as the central pillar of an employee’s day (and therefore their productivity and engagement).

DEXOps operationalizes experience by starting with the immediate concerns and frustrations of individual business units, understanding what the consequences of these issues are, and finding technological solutions that eliminate the root causes of digital friction.

Rather than focusing on technical capabilities, IT leaders should start by asking other business units:

What are your team’s top priorities?

Where are you experiencing friction or risk?

What does success look like for you and how is that measured?

Do you know what value you bring to the overall business?

It’s amazing how much more engaged other parts of the business will be when, instead of diving into ISO compliance requirements, the conversations revolve around how their most pressing challenges can be fixed.

XLAs are an opportunity as well as a challenge

Many businesses immediately tend to view new regulation as a burden. But ISO 20000-18 and the increasing role of XLAs also represent an opportunity to accelerate the changes already happening to the role of the IT department. The shift from infrastructure to experience is tricky and requires businesses to be open to significant realignment across traditional roles, responsibilities, and metrics.

Some organizations will continue to ignore the importance of experience, but the broader changes that have driven the drafting and implementation of the regulations are here to stay. Businesses that see the new standard as an opportunity to use regulation to create a competitive advantage will have a crucial head start on those that are still stuck in the old, outdated, reflexive IT model.

We've featured the best business intelligence platform.

This article was produced as part of TechRadar Pro Perspectives, our channel to feature the best and brightest minds in the technology industry today.

The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/pro/perspectives-how-to-submit

Over the past year, I've had countless conversations with treasury and payment teams at major banks, and I'm often asked the same question: how do we compete with wallet-native fintechs on cross-border payments? But as I tell these teams, this isn't the right question to be asking anymore. Banks no longer need to play catch-up; now they can play to win.

It starts with Swift's payments scheme: the rulebook, built with more than sixty banks, that commits participants to full-value delivery, upfront fee and FX transparency, and end-to-end tracking on international transfers. It's already live across major corridors.

And now the real-time domestic rails banks already use to move money instantly within a country are being extended to handle the cross-border leg of a transaction too. Together, banks can extend their reach into digital wallets across the globe by connecting to networks that already exist, rather than building new ones from scratch.

That reach matters well beyond the traditional remittance corridors most people still picture when they think about cross-border payments. While global remittances represent huge sums of money ($905 billion in 2024, according to World Bank estimates – up 4.6% on the previous year), they are no longer the whole picture.

Increasingly, the volumes crossing borders come from businesses paying each other, platforms paying creators and freelancers, and small firms invoicing overseas customers directly. Every one of those flows still needs to arrive somewhere the customer trusts, and that is where banks already hold five advantages that fintechs cannot easily replicate.

Rather than worrying about being outmaneuvered by fintechs, I challenge you to think about the advantages you already have. So, here's a better question: which of these will you put to work first?

The account relationship starts and ends with you

Every cross-border payment revolves into an account somewhere, and for the majority of consumer and business transactions, that account sits with a bank. Fintechs and wallet providers can originate volume, but at some point, in the chain, funds still land in a regulated deposit account banks already hold.

What has changed is the ability to extend that advantage outward: connecting existing account infrastructure directly to wallets and local schemes elsewhere, rather than routing through a chain of correspondents to get there.

Holding the funds is what holding trust means

In the customer experience at a bank, whoever holds the funds determines whether a payment arrives intact or arrives short, deduction by deduction, across a chain of intermediaries.

Swift's payments scheme commits participants to delivering the full value of a payment with no deductions along the way. Banks are the natural home for that commitment because they already hold the funds under direct regulatory oversight, rather than passing them through intermediaries without transparency.

Supervision and regulatory oversight are what customers are already paying for

Banks operate inside some of the most heavily supervised frameworks in financial services, and that supervision is precisely what customers are buying when they choose a bank for a cross-border payment.

Combined with the move to ISO 20022 messaging, now the mandatory standard for cross-border payment instructions, which carries richer, more structured data through CBPR+, banks can now show customers more of what is happening to a payment in transit, not less, under the FATF travel rule being applied.

No fintech or wallet provider can replicate the confidence that comes from being the regulated counterparty in the chain and the enforceability of the FATF travel rule.

Trust built over decades cannot be matched and built in months

For high-value corporate and SME flows, trust is the existing product offering, and it does not transfer easily to a newer entrant, however good its infrastructure. That matters more, not less, as the flows crossing borders diversify: SMEs trading directly with overseas buyers, gig workers and independent creators paid from platforms based in another country, and B2B and B2C commerce settling at volumes few payment providers were built to handle a decade ago.

Every one of those flows still needs a bank willing to stand behind it. Banks that have already spent decades building that trust are better placed to extend it into these newer categories than to concede them.

Only the bank can control and own the whole journey, start to finish

Real-time domestic rails are now reaching for the border. The Clearing House's RTP network, for example, is targeting September 2026 to support correspondent bank activity for one-leg-out international payments, giving banks a way to originate the domestic leg of a cross-border transaction on infrastructure they already use daily.

Bank of America has already announced a cross-border real-time payments service built on Swift and its own CashPro platform. Neither move required building new global rails from scratch. Both give the bank full visibility and control from origination to final credit, a position no correspondent chain or third-party wallet provider can offer in the same way.

None of this is guaranteed, and none of it is evenly distributed. Community and regional banks without the scale of a Bank of America face a real cost to adopting real-time and ISO 20022-native infrastructure, and regulatory harmonization across corridors remains uneven even as individual schemes go live.

The advantage described here is available to banks that choose to invest in using it. It is not conferred automatically by size or incumbency alone.

Banks already have most of what this moment requires: the accounts, the funds, the regulation, the trust, and now the connectivity to reach beyond their own rails. What happens next depends on who decides to use it.

The opportunity is clear: Banks have all the foundational strengths to capture the rapidly growing real-time cross border payments-to-wallets market. By combining their established customer base with modern payment networks with the SWIFT scheme solution, banks can transform cross-border payments into a faster, more accessible, compliant, and customer-centric experience.

We've featured the best money transfer app.

This article was produced as part of TechRadar Pro Perspectives, our channel to feature the best and brightest minds in the technology industry today.

The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/pro/perspectives-how-to-submit

Autonomous AI agents are already running inside core infrastructure – executing code, applying policies, and managing DevOps functions. And the projects keep stalling, because the security models they’re being wired into were built for a world that no longer exists.

Retrofitting non-deterministic actors into those models is costing engineers time they don’t have, and it introduces risk no enterprise can manage.

Many projects have stalled amid concerns about deploying without a robust security foundation – and with good reason. We’ve already seen an agent delete a company’s entire production database, and its backups, in nine seconds. Security teams are being asked to stop scenarios like that with tools built for a world of two actors. The cracks are starting to show. Something has to change, or innovation stalls under the weight of its own controls.

AI agents require a new identity model

The pressure on production and engineering teams to speed up delivery is very real and pervasive. So they often fall back on old habits like granting agents broad privileges and treating them as any other microservice.

But agents are very different from machines; they are error-prone and non-deterministic, just like humans. Yet, operating at machine speed, 24/7. Agents can delete entire production databases in nine seconds. How many humans do you know who could do that?

And this brings me to the crux of the problem. Identity systems were built for a world with two kinds of actors – humans and machines – but there are now three. Trying to fit agentic AI into outdated systems makes each agent a potential source of compromise, and one that can execute thousands of actions across infrastructure in seconds.

Yet this is what engineers are asked to do; stop catastrophic scenarios with legacy IAM tools that are breaking down. The cracks are starting to show.

Why the old model breaks

Historically, identity fragmentation has plagued engineers working with Kubernetes clusters, cloud platforms, container orchestration, CI/CD pipelines, databases, etc.

For a human workforce, this was manageable. Humans are trackable; they log in and log out. They are slow enough that visibility gaps rarely turn into immediate incidents.

Enter agentic AI, and the speed gets turned up to the max. Suddenly, teams are inundated with thousands of activity logs, and they lack the ability to effectively contain the agent before it executes unauthorized changes.

Trying to enforce strong authentication and short-lived privileges would mean building individual integrations for every tool in the stack. It makes AI hard to scale when each tool uses a different integration protocol.

Rather than focusing on innovating with agentic AI, engineers are forced to stitch together IAM, infrastructure, and secrets by hand — with no consistent identity, no visibility into agent actions, and every team building its own container or VM workflows from scratch.

But creating a new tool to handle a third identity type is the worst reaction the industry could have. It would double the work for engineers, as they would need to rebuild the identity policy from the ground up and introduce greater anonymity. A new identity silo is anonymous to other siloed systems, making it even harder to catch attackers.

This leaves us with the question of how enterprises can control an agent's behavior. The solution isn’t about adding to the tech stack or implementing more tools; it's about changing our identity models to eliminate anonymity entirely.

AI control means zero anonymity

To remove anonymity from infrastructure, enterprises must give every actor –spanning humans, machines, workloads, and AI agents – first-class identities, cryptographically secured by a hardware root of trust.

Ditch static credentials entirely. Eradicating API keys and passwords eliminates the credential sprawl that causes breaches, as well as the threat of secrets being stolen or handed over to the wrong actors. With identity rooted in real-world factors, attackers cannot impersonate a trusted machine and trick an agent into exfiltrating a database.

But strong authentication in itself is not enough. AI agents, like all other actors, need to adhere to zero-trust principles. This can only happen when siloed systems are replaced by an infrastructure layer in which agents have the exact same identity type as the machines they run on and the humans who authorize them.

Agents should operate with short-lived privileges tied directly to specific actions authorized by a human user. Privilege attached to the action, not the actor. For example, an agent generating code must inherit its mandate from a human owner with matching authority, restricting privileges to only the specific data tables required for that task.

Non-deterministic actors also require a contained, trusted execution environment before touching production infrastructure.

With no default privileges, the blast radius is heavily bounded. But this can only be achieved when a single policy is set by a single system for all identities.

Identity policy can also be introduced as an enforcement layer between the agent and its inference endpoint, so behavior is controlled before instructions are ever executed.

With a unified architecture, identity becomes the control plane for AI

AI agents are unlocking immense opportunities in enterprise environments, especially when deployed in live infrastructure, where they deliver the most value. From managing routine changes to fixing deployments in real-time, the possibilities are endless. Succeeding with AI in such critical business operations requires tight control of behavior.

We've featured the best AI tool.

This article was produced as part of TechRadar Pro Perspectives, our channel to feature the best and brightest minds in the technology industry today.

The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/pro/perspectives-how-to-submit

By almost any measure, Tim Cook has been one of the best tech CEOs ever. He diversified product lines, launched new lines of business, expanded across the globe, and grew Apple's overall market value ten-fold.

Cook is the rare public tech figure who has, for the most part, survived the vagaries of a public life. Did he disappoint many when he appeared to bend the knee before the United States' divisive President? Absolutely, though if you understand Cook, and you listen to his explanations, you know why he did it and how it actually fits into his worldview. Generally, though, Cook's been on a 15-year-long winning streak that was made possible by both sticking to the strict ethos of Steve Jobs' original Apple charter and by pragmatically doing things his own way.

I once wrote that Steve "Jobs was Apple." Tim Cook made Apple. Where Steve Jobs' mercurial personality was intertwined with Apple across its products, development, and marketing, Cook's fingerprints are so absent from the company he's run that you might wonder if he has them.

Cook became a celebrity at Apple, but without the rabid hero worship Jobs would sometimes foster.

I'd argue, though, that in an era of a million rising individual influencer personalities, Cook may have been exactly the CEO Apple needed. He commanded the stage at keynotes and even engaged in some playful "acting" for short comedic product segments, but Cook has clearly never been interested in building on a cult of personality. When the opportunity to speak in his own biography arose, Cook characteristically deferred to the people who know him. As a result, his voice is largely absent from a tome purporting to tell the story of his remarkable life.

Cook built it to last

What Tim Cook has done is build an Apple for the ages. At his fifth anniversary, I wrote about how often he proved "he was about choice." It wasn't a complete departure from Steve Jobs, who diversified the iPod line, but Cook carried that idea through virtually every product category. In the iPhone, this has provided the freedom to offer not just a wide range of iPhones, but handsets at a wider range of sizes and prices, opening up the cachet-filled handset to an even bigger audience.

Cook took Steve Jobs' germ of an idea for "technology that would improve the relationship between patients and health care providers," and turned it into the Apple Watch. Over the next decade, Cook and company would lean into that ideal, continuously enhancing the wearable to turn it into a true life-enhancing fitness and health wearable. There was obvious pride in Cook's voice whenever he shared true stories of the Apple Watch detecting potential health issues and possibly saving lives.

One of Cook's most remarkable creations is the Services business, originally built around a collection of "Plus"-named services that include Apple TV+ (now just Apple TV), Fitness+, Apple News, iCloud, and Apple Music. Services is now Apple's fastest-growing business, recently reaching over $30 billion in revenue in the last quarter.

I remember sitting in the Steve Jobs Theater audience in 2019 as Cook introduced a launch lineup that included Jason Momoa, Oprah, and Big Bird, wondering how, even with all that star power, Apple TV+ could compete with Netflix. Today, that's almost a laughable concern, as Apple TV enjoys some of the most acclaimed and widely-watched shows in streaming, including Widow's Bay and Ted Lasso. With movies like F1, Cook has even built a formidable movie studio.

It's not that I don't think Steve Jobs couldn't have built something similar. After all, it's unlikely we'd have Pixar's Toy Story franchise if it weren't for Jobs, who helped pivot the software company to filmmaking. Still, I think Jobs saw the Tech and Entertainment industries as two parallel but never intersecting lines. Cook's head was always on a swivel, and he has focused on business-strengthening diversification.

Missteps and power moves

For Apple purists, services might be a less exciting and interesting area than new product categories like the aforementioned Apple Watch, and Cook's track record in pure-play innovation has been somewhat spotty.

Cook's biggest swing, the mixed-reality Apple Vision Pro, remains one of the most impressive pieces of technology I have ever seen, but I have to wonder if Jobs would've made the same call, shipping a $3,499 headset and claiming it was for consumers. I have a strong feeling that the Vision Pro will not survive deeply into the John Ternus era.

Similarly, Apple under Cook has fumbled the AI ball, a fact that Apple has more or less admitted to us. It remains to be seen if the far more intelligent Siri AI will deliver on all the Apple AI promises and finally put it properly into an Artificial Intelligence race that has all but passed it by.

These are minor blemishes in an otherwise sterling track record, one that may ultimately be defined by the stunning transformation from a collection of Mac hardware that relied on Intel CPUs to an impressive array of Apple Silicon systems that perform well at virtually every level of demand.

If there's a natural handoff from the efforts of the outgoing CEO to the incoming one, John Ternus, it's this one. As a longtime hardware lead, Ternus has been deeply involved in all Mac innovations. The first time I spoke to him, it was about the short-lived 12-inch MacBook and all the work Apple did to move antennas and speakers and create a layered battery system. More recently, Ternus fronted the impressive MacBook Neo launch.

True continuity

Unlike Cook, who took over for a dying and idolized CEO in 2011, Ternus steps into the seat of a moving car as the deeply respected Cook scoots over to an Executive Chairman role.

15 years ago, people openly questioned Apple's ability to continue in Jobs' absence (I did not, writing, "Apple will be fine without Steve Jobs"). That concern does not exist in this transition. Ternus not only knows and understands the strategic plans laid out by Cook, but he can also still turn to his mentor at any moment for additional guidance.

As was the case when Cook took over, there won't be many changes evident in the first 12 months of Ternus' control. But a decade from now, there will be new products and services, devices sunsetted, and endless post-mortems about how well or poorly Ternus handled the Siri AI rollout.

Tim Cook was no Steve Jobs, which turned out better than fine for Apple. Similarly, Ternus is no Tim Cook. They have different temperaments and skill sets (Ternus is more of a true technologist). In the end, this next era of Apple is another exciting mystery full of anticipated unveilings like a folding iPhone and AR glasses, and surprises like Apple robots and EV cars (equal chance those don't happen), but the biggest surprise of all might, as it was with Cook, be what the company will be like under Ternus. That, and how Ternus will greet us all when he takes the stage for the next iPhone launch on September 9. Maybe, "Hey, everyone"?

For anyone working in IT and familiar with SaaS sprawl, the idea we’ll soon end up with agent sprawl may set off alarm bells. Agent sprawl refers to the rapid, uncoordinated deployment of AI agents across departments, each built on different models, governed by different rules, and often disconnected from core business workflows.

But while SaaS sprawl caught enterprises off guard, this time organizations have no excuse, and need to put measures in place now to avoid repeating the same mistakes. However, right now, most organizations are building agents in isolation and optimizing for local productivity rather than enterprise value.

Thankfully, with the right control layer funneling every agent through the same governance and security standards within an organization, the agent-powered future can be far safer, scalable and simple. 

Agent sprawl is inevitable

Just like SaaS before it, agent sprawl is already happening. Each department within organizations is experimenting with different agent technologies to address specific challenges or work more efficiently.

A sales team may deploy a CRM-based AI assistant to qualify leads, while developers build their own coding agents, and marketing adopts a separate content generation tool. Each delivers local value, but none are aligned or connected to the other, creating a familiar-looking sprawl.

This fragmentation introduces a range of challenges that will only increase over time. With no single owner responsible for how agents are deployed or monitored, there’s a lack of governance and oversight. Security risks increase as agents gain access to sensitive systems without consistent controls.

Teams may unknowingly duplicate efforts, solving the same problems in parallel with different tools. At the same time, many agents are deployed without clear links to business outcomes or understanding the wider context, making it difficult to measure return on investment and distinguish meaningful innovation from experimentation. Lots of AI tools may be exciting, but deliver far more style than substance.

Is sprawl inevitable? Yes, probably. So the question now becomes, how do we control it?

The role of orchestration

Having an orchestration layer can bring order to this complexity by creating consistency in how agents are governed, secured and deployed across the business. For SaaS applications, orchestration provides visibility into what tools already exist and gives a clear view of enterprise workflows.

With agents, it outlines what agents are already available and how they operate, while also controlling how they access data, tools and existing technology. Orchestration is the critical element that turns agents from experiments into scalable business infrastructure, for example, assigning agents to business workflows and measuring their performance against real outcomes.

Too often, we see enterprises experimenting with different agents that have no connection to the core business processes and goals, meaning they fail to deliver any ROI. For example, some portions of a business simply do not require AI technology and will unlikely see benefits in the same way as another department. Orchestration can unearth and filter out agents that are unnecessary, saving businesses money. 

Building trust in AI-driven enterprise

Rather than trusting individual agents, organizations should focus on trusting the system that governs them.

This is where orchestration becomes a trust layer, ensuring every agent operates within clearly defined boundaries with consistent oversight and accountability. Instead of individual teams managing risk in siloes, organizations can centralize control while still enabling innovation across different departments and business functions.

In the same way orchestration brought order to SaaS sprawl, it can do the same for the next wave of enterprise AI. It brings the same ease, security and integration with existing SaaS tools, while ensuring only high-value agents make it into production, supporting innovation and keeping within the business goals and purpose.

As agents become embedded across every function, orchestration will shift from a technical layer to a core enterprise capability. The winners will be those who can control, connect and trust their agents at scale.

We've featured the best AI website builder.

This article was produced as part of TechRadar Pro Perspectives, our channel to feature the best and brightest minds in the technology industry today.

The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/pro/perspectives-how-to-submit

For years, enterprise Software as a Service (Saas) has enjoyed a comfortable position at the center of organizational processes. Users who wanted to raise a ticket, process an invoice or manage a workflow would log into their employer’s cloud application. The user interface was the gateway to the work, and the provider’s business model followed naturally: charge for seats, expand adoption, and grow revenue as more users spend more time inside the platform.

Agentic AI is beginning to change that. As users increasingly deploy AI agents, either to carry out tasks within SaaS platforms or to deliver functionality outside them, organizations require fewer seats – hitting providers’ subscription revenues.

At the same time, AI is making bespoke software faster and cheaper to build, giving organizations a credible alternative to buying another SaaS platform. Why accept subscription costs and vendor lock-in, when cheaper alternatives may achieve the same outcome?

None of this means SaaS is about to disappear, but its position at the center of enterprise software is under threat. SaaS vendors are facing three structural challenges that will reshape the economics of enterprise software.

Agents become the user

The first threat is simple: users may have AI agents carry out tasks within the SaaS platform, reducing the need for human users, or build agentic capabilities outside those platforms altogether.

That is a serious problem for any SaaS vendor whose business depends on people living in the interface. If an agent can resolve a support query, update a record or complete a routine transaction, the user never has to touch the application.

We’re already seeing this in practice. In some IT service management projects, agents now resolve such a high proportion of tickets that organizations need far fewer users interacting with the SaaS platform. As buyers begin questioning whether they still need premium software for the shrinking proportion of manual work, the economics start to shift.

Bespoke is back

The second threat is the one SaaS vendors dislike most: buyers may once again be willing to build new capabilities from scratch.

For the last decade or so, ‘build versus buy’ was often a non-debate. Custom software was too slow, too expensive and too risky, so SaaS usually won.

AI is starting to shift those calculations. As software development becomes faster, cheaper and more adaptive, some solutions that previously defaulted to SaaS start to look buildable – especially when the organization needs a narrow set of capabilities wrapped around its own processes and data. The more customization a SaaS platform requires, the weaker its cost-efficiency and automated-upgrade advantages become.

This does not mean large enterprises will abandon mature SaaS platforms overnight. Their resilience, integrations and compliance capabilities remain valuable. But once customers believe they have a credible alternative, vendors lose pricing power and the assumption that renewal is the obvious decision

Business model redundancy

The third threat is that even when the product survives, the commercial model may not.

Traditional SaaS pricing was built for a world in which value correlated reasonably well with the number of people sitting in the application. That logic gets shaky when agents perform work that would once have required more licensed users, and more operational dependency on the interface.

Vendors can see this. Many are shifting towards consumption-based pricing, exposing services directly to agents or redesigning products for an agentic world. The common thread is clear: once software starts working for people rather than being operated by them, seat-based economics begin to look outdated.

From the customer’s perspective, that could be a good thing. Paying for usage is often cheaper than buying licenses. The challenge for vendors will be proving they’re genuinely adapting to an agentic future, rather than simply repackaging existing products to protect margins.

SaaS is not dead

To be clear, this is not a “SaaS is over” argument. Many enterprise platforms will remain valuable, providing governance, resilience, compliance, and years of accumulated functionality that are difficult and expensive to recreate.

What will change is where value sits. In an agentic world, SaaS will be less the primary interface through which work gets done; more a component in a broader ecosystem of agents, orchestration and systems of record. Some platforms will adapt and become stronger; others will discover that the value they once captured has shifted elsewhere.

As for the corporate users of SaaS, the best advice we can give is not to wait for perfect conditions before acting. No business ever reaches a point where its data, processes or architecture are “finished”, and SaaS providers will continue to adapt their offers.

Meanwhile, experiment, learn and adapt in your delivery of administrative and corporate functions: in some functions you may find that agents drive down SaaS costs; in others, greenfield development may now be viable. In still other cases, adding new functionality using agents that leverage SaaS solutions but sit outside those ecosystems keep future options open.

So do something that works for you now – but keep the option of moving again in a year or two.

We've featured the best AI tool.

This article was produced as part of TechRadar Pro Perspectives, our channel to feature the best and brightest minds in the technology industry today.

The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/pro/perspectives-how-to-submit

The most damaging AI-related security incident your organization faces this year probably won't originate from external attackers using sophisticated new models. It's far more likely to begin with an employee pasting a client contract, a financial forecast, or a set of HR records into an AI tool because it makes their job easier and nobody has told them why it matters.

Shadow AI is a growing problem, and our research found that nearly half of employees at larger enterprises regularly feed corporate data into AI tools that nobody in IT has approved or governs.

More striking still, 85% of employees continue doing so even when company-sanctioned tools are available, pointing to a governance failure that runs all the way to the executive suite. With shadow AI, sensitive data can move silently outward through channels most security stacks were never designed to intercept.

Adding to the security risk of this unmonitored data flow, the advent of the EU AI Act also means organizations now face specific legal demands on managing AI use. The ability to have full governance over how AI is deployed, governed and monitored, is becoming a regulatory, as well as a security, imperative.

Why the EU AI Act makes this a board-level problem

Shadow AI represents a serious security issue, with IBM's 2026 Cost of a Data Breach report estimating that unauthorized tools contributed to 43% of breaches over the last year Now, the EU AI Act is adding significant regulatory requirements on top of these risks.

The Act's obligations have rolled out in phases; most recently, organizations classified as general deployers of AI have new inventory, data governance, audit logging and transparency obligations as of 2nd August 2026.

Other deadlines have shifted further ahead, with controls over high-risk AI usage, covering areas like recruitment, credit scoring and biometric categorization, set to come into force from 2nd December 2027. AI embedded in regulated products will be covered from 2nd August 2028.

Any organization whose employees use AI systems now has compliance obligations as a deployer, regardless of whether those systems were formally sanctioned.

All organizations using AI should be aware that the AI literacy obligation under Article 4 has been enforceable since February 2025, meaning organizations are on the hook for ensuring their employees are aware of safe and sanctioned AI use.

Rules around high-risk AI usage will also apply to more operations than it may seem at first, including an employee using an unapproved consumer tool for tasks like screening CVs, assessing creditworthiness, and evaluating performance.

These are common tasks that could trigger the full weight of the Act's oversight against a system that the IT department didn’t even know had been deployed.

With penalties reaching up to €15 million or 3% of global annual turnover for high-risk breaches, many organizations are carrying more exposure than they realize.

Why your existing security stack can't see it

The challenge with shadow AI is that it exploits the blind spots between conventional security layers, slipping through gaps that most tools were never designed to close.

CASBs and secure web gateways cannot decrypt conversational data flowing to legitimate LLM domains over HTTPS, for example. From the network's perspective, a prompt containing a full customer database is indistinguishable from any other encrypted web session. Browser extensions are limited to managed endpoints, blind to personal devices and AI embedded within approved SaaS.

Likewise, API gateways are usually built around authorized enterprise deployments, which means they capture the AI activity organizations have already approved while missing the consumer-grade AI tools driving most risk.

These blind spots compound with each other, so an organization running all three layers may still have no visibility into AI activity across a significant portion of its estate, and no means of generating the interaction logs or policy enforcement evidence the Act requires.

What good AI detection looks like

Controlling AI data flows is usually managed by workers performing their duties without malicious intent. However, it’s remarkably similar to defending against an external threat actor covertly accessing your data.

The detection logic needs to match that reality. Endpoint-native detection intercepts sensitive data at the point of movement before it reaches an external AI system, enforcing policy at the prompt level across managed and unmanaged browsers, personal devices, and AI functionality embedded within SaaS tools. It operates where the activity occurs, rather than attempting to catch it downstream.

Given the scale of the potential fines, the ability to prove compliance matters almost as much as preventing security breaches.

Continuous discovery across the estate, including any unsanctioned tools gives organizations the AI system inventory the Act requires.

Granular interaction logs - who used what, when, and what data was involved - satisfy the documentation requirements under Articles 12 and 13, without teams needing to scramble to reconstruct activity after the fact. The same data pinpoints exactly where AI literacy gaps exist, making Article 4 compliance something demonstrable rather than simply asserted.

Practical steps for compliance

Security and compliance teams aiming to comply with the EU AI Act have a clear path to follow.

The starting point is mapping the full AI estate. Discovery needs to extend beyond IT-approved tools to unmanaged endpoints, personal devices on corporate networks, and AI embedded within SaaS.

Data governance must move to the endpoint. Policies prohibiting sensitive data sharing with unapproved tools are not technical controls and, by the time enforcement happens at the network edge, the data has already left.

It’s important to remember that information shared with an external AI system may be retained in prompt logs, incorporated into model training data, or held on servers in jurisdictions the organization has no visibility into. The moment data crosses that boundary, the organization loses control of it entirely, and no policy document will retrieve it.

And since Article 12 requires interaction records that can be handed to regulators on demand, organizations will need to have continuous, automatic auditing of both activity and security measures.

Finally, AI literacy programs aimed at improving user awareness should be driven by behavioral data rather than generic training programs. Activity logs showing where governance failures are occurring - at senior leadership level as much as anywhere else - provide both the diagnosis of the issue and the evidence regulators will want to see.

We've featured the best antivirus software.

This article was produced as part of TechRadar Pro Perspectives, our channel to feature the best and brightest minds in the technology industry today.

The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/pro/perspectives-how-to-submit

The numbers circulating through fintech right now are hard to ignore. Almost 10,000 jobs have been cut across the global sector in 2026 so far. PayPal is eliminating roughly 20% of its workforce. Block has shrunk to fewer than 6,000 employees, and Coinbase cut 14% of its headcount.

And hovering over all of it is a single explanation, repeated so often that it starts to remind us of a dot-com bubble — artificial intelligence. But is that actually what's happening? And more importantly, does it matter if the story isn't entirely true?

When the story becomes the strategy

AI accounted for 40% of all announced layoffs in May 2026, up from much lower levels earlier in the year. At first glance, it seems like a natural explanation: if the technology handles the routine work better than a human, then why not cut costs?

But the thing is, when we start looking closer at why exactly these mass layoffs are taking place, a deeper structural problem appears. So far, AI is just a good justification for how the industry communicates these shifts.

Deutsche Bank analysts identified this neat disguise in January. They called it "AI redundancy washing." Put simply, it’s the practice of attributing workforce reductions to AI adoption when the underlying drivers are far more familiar.

But if it’s not really AI which replaces the workers, then what are the real reasons for all these layoffs? If each company were to be considered individually, it could be anything, from deteriorating unit economics to the pressure of demonstrating operational discipline after years of “growth-at-all-costs.”

Research points to the same explanation: nearly 60% of US hiring managers acknowledged that they emphasize AI's role in layoffs just because it lands better with stakeholders. Indeed, a headcount reduction could have sparked fears about companies' growth models, but framing it as a strategic AI implementation makes it sound good.

The productivity paradox

Let’s assume that some of the companies are indeed replacing their workers with AI tools. In this case, there should be numbers to show it. But they don't share any effectiveness ratios, at least not yet.

A survey of 350 executives at companies with over $1 billion in revenue found that 80% of organizations piloting AI or autonomous technology reported workforce reductions. But after that, many honestly admit that they cut jobs regardless of whether the technology was actually generating returns.

This way, planning to spend another $10 million or more in 2026, companies are betting on a future that hasn't fully arrived. And trimming headcount in anticipation of a productivity windfall that remains, for most, theoretical.

Even if AI is implemented in a particular organization, it’s mostly in pilot mode, sitting on top of existing infrastructure rather than remaking it. Moreover, the AI payback picture is modest: only 25% of AI initiatives deliver the expected returns, and only 16% reach the scale of the entire organization.

McKinsey’s 2025 global AI survey found the same results. While most firms are using AI in at least one function, nearly 70% are still stuck in experimentation or pilot stages rather than scaling AI across the enterprise. And what’s more upsetting is that no more than 39% reported a significant EBIT impact from the technology adoption.

Where the capital is actually going

If AI isn't yet delivering the productivity revolution that many executives promised, then where is the industry's attention? The answer is less exciting than the headlines suggest — only last year, infrastructure and internal business systems received more than $200 billion worth of investments.

For now, fintech companies are under pressure to prove that all the promises made around AI weren't just another cycle of hype. So instead of reinventing financial services overnight, many are turning inward, looking for ways to make existing operations cheaper and easier to scale.

Even though the AI hasn’t brought the desired effect yet, the shift to automation is still important. For the past decade, fintech has largely been focused on the customer experience to offer a disruptive product, but from now on, more effort will go into the mechanisms behind the product than the product itself.

But this doesn’t mean that the "AI replaced the workers" narrative feels complete. Cutting headcount and replacing employees with bots may look good on a quarterly report, but it's not much of a strategy. Cost reduction can improve margins in the short term, but it rarely creates lasting differentiation.

So, in this environment, the winning ones will be those asking and answering honestly to a different question: What can we offer the market today that we couldn't offer a year ago?

The value of AI innovation beyond cost-cutting

Arguably, the more important part of AI adoption isn’t about how it replaces workers at all, but rather about what new value it creates. Historically, the biggest technological shifts rarely created value by simply making existing processes cheaper. They created entirely new products, services, and business models that were impossible before.

Without going too far in search of examples, let’s just remember the Internet. When it first arrived, a lot of people justified investing in it by bringing up “efficiency” aspects, such as digitizing paperwork or reducing communications costs through e-mails.

In reality, though, the Internet went much further than that, creating entirely new business models: online media, social networks, search engines, cloud software, and more. Giants like Google or Amazon did not get to where they are now simply because they optimized their processes. They found success precisely because they built products that simply couldn't have found success before the Internet became widely adopted.

Now, it’s AI’s turn to go down that road. What innovations it will bring is a story that still remains largely unexplored in fintech, but the potential is massive. In this sense, today’s hyper-focus on workforce cuts is likely going to be just a short phase in a much larger legacy that AI will leave in finance. The true winning move for companies now is to work out how they can use it to better serve their customers in new, never-before-explored ways.

Cost savings may be a measurable and immediate metric, but over time, the true advantage of adopting this technology will come from what it brings to the table, not what it takes away from it.

We've featured the best personal finance software.

This article was produced as part of TechRadar Pro Perspectives, our channel to feature the best and brightest minds in the technology industry today.

The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/pro/perspectives-how-to-submit

You can spend hours choosing the best laptop, comparing processor options with the display and wondering how long the battery will last.

Then you switch it on and discover another part of the experience that rarely enters the buying decision: the manufacturer’s preinstalled software, often billed as tools to help optimize your device.

Sometimes, they earn their right to come pre-installed. The pre-installed software might be where you change performance mode or cap the battery charge, alongside model-specific settings Windows 11 simply doesn’t offer.

The strange part? How much bigger the role of pre-installed software (sometimes called 'bloatware') has become. A simple utility can now feel like another software layer sitting on top of Windows 11, with launchers and user accounts jostling with the settings you actually want.

And because some of those controls are legitimately important, uninstalling the whole thing isn’t always an option.

Windows 11 can’t actually control everything on your Windows laptop

For all the settings tucked inside Windows 11, there are still plenty of things it can’t manage on its own.

And this isn’t just a gaming laptop quirk. On supported Asus models, MyASUS handles things such as Battery Care Mode and model-specific fan profiles that Windows 11 doesn’t offer.

Gaming laptops simply take the idea much further.

Take the Alienware 16X Aurora. Alienware Command Center offers several performance settings, and in our testing, its 'Overdrive' mode pushed the fans harder while making use of the laptop’s Cryo-Chamber cooling system.

None of that feels especially unreasonable.

Laptop makers know how their own hardware is designed to behave, and a dedicated utility gives them somewhere to offer controls that would otherwise be buried in the BIOS or unavailable altogether.

The problem starts when that useful little control panel begins acquiring ambitions of its own.

A settings menu somehow became an entire platform

Many of these apps now go well beyond looking after the laptop itself.

Armoury Crate still gives Asus ROG owners useful controls over performance and cooling, but it also has a user account system and optional content services alongside its device settings.

Lenovo’s Legion Space pushes the idea further, putting thermal controls alongside a game store and its AI-powered Gaming Zone. On supported machines, HP’s Omen Gaming Hub mixes fan and graphics controls with a game library and GeForce Now.

Taken one by one, most of this is easy to defend. Put it all in the same app, though, and the hardware settings start to feel like one part of a much larger product.

If I open software to quieten the fans, for example, I’m probably not also looking for a new game or a giveaway.

The worst part is that you actually need it

With ordinary pre-installed software, there’s an obvious solution if you don’t want it: uninstall it. Laptop software made by the manufacturers themselves is harder to shake because some settings can make a real difference in how the machine behaves.

Our MSI Raider 16 Max HX review shows why this gets awkward. The laptop comes with a fair amount of bundled software, yet MSI Center itself gives you useful access to performance presets and RGB controls.

Wiping out every manufacturer utility for a cleaner Windows 11 install can take useful features with it. However, some laptops do have alternatives.

G-Helper is an open-source option for Asus machines, offering access to performance and GPU modes, battery charge limits, and other hardware controls in a lightweight app. Its built-in performance modes use the same firmware presets as Armoury Crate.

You don’t have to be comfortable swapping official software for a community-made utility to see the appeal. G-Helper’s existence shows that granular laptop controls can live in a focused tool rather than a whole, often bloated, ecosystem.

Microsoft bears some responsibility too

Some of this starts with gaps in Windows 11 itself. Smart Charging exists on compatible devices, for example, but there’s no universal control for choosing a fixed charge limit – Microsoft says each manufacturer implements the feature differently.

Microsoft has already shown what a more standardised approach can look like.

Dynamic Lighting is a good example: compatible RGB hardware can now be controlled directly through Windows 11, rather than relying entirely on separate manufacturer utilities.

Fan tuning and firmware are much more specific to each machine, so manufacturer software isn’t going away.

But every common control Windows 11 can standardize is one less reason a laptop needs its own parallel settings software - and that would mean less bloat preinstalled on our devices.

Laptop software needs to become boring again

Maybe the real problem is that these apps have become something you even need to think about. A good pre-installed app should make the laptop easier to live with, then stay out of the way until you need it again.

MSI Center at least gets some things right: it’s modular, so you can install only the features you want rather than loading everything by default.

Switch brands, though, and even familiar jobs can move somewhere completely different. A 'quiet' fan mode that takes seconds to find on one laptop might be buried under different terminology on the next.

Software like this probably deserves more consideration in laptop reviews, too. It won’t rescue poor battery life or a bad keyboard, but if you rely on it to change how the laptop behaves, it's part of the product you bought.

I don’t want laptop makers to stop building these tools. I just want them to remember what they’re for: helping me use the laptop I already bought.

Although AI is advancing at a rapid pace, the idea that it might one day replace humanity has thus far been consigned to the realm of fiction.

Rather than signalling the end of mankind, however, others have speculated that an advanced AI may one day simply supplant our position at the top of the food chain and live harmoniously with us, a species that will live to serve it.

Electric sheep

The American mathematician and computer scientist Claude Shannon was giving an interview with Omni Magazine on his work and legacy when he speculated on the role of AI and robotics in the future.

His comments came during a conversation about the limits of digital computers and the piping that backs so much of the AI technology we use today, artificial neural networks. He suggested, despite the sour mood around AI at the time, that it would be one day so advanced that our collective intelligence as a species would pale in comparison.

Far from being a doomsday scenario, though, his likening of humanity to a dog indicated a sense of companionship, mutual love and respect, and a sense of loyalty that would cement mankind's role in a world dominated by a future machine intelligence.

Roboslop

At a time when funding for AI was running dry and we encountered many bottlenecks like the lack of compute power, science fiction writers and scientists alike both started envisioning ways in which future systems could one day replace us.

Although none of those predictions have come to pass, there is a widening consensus that one day the technology will grow to surpass the level of human cognition.

This hypothetical technology, known as artificial general intelligence (AGI), may not be invented any time soon – and we may not presently have the architecture required to see progress. But scientists expect this to come within living memory.

Once we reach this point, which many refer to as the technological singularity, it's wholly unclear what the implications may be and whether AGI may eventually give rise to a cascading series of more intelligent systems that culminate in an artificial superintelligence (ASI).

Nvidia's DLSS 5 has been nothing short of a bubbling cauldron of controversy ever since the next-gen gaming tech was first aired. That initial reception shocked Nvidia as what seemed like a legion of PC gamers took up arms and vowed, well, all sorts of colorful things — most of them involving the term 'AI slop' in some fashion.

However, there are signs that this overwhelmingly negative tide of opinion is turning courtesy of a fresh revelation this week. This comes in the form of a leak rather than anything official from Nvidia, with Team Green's only DLSS revelations at Gamescom 2026 pertaining to the current incarnation of the feature (namely DLSS 4.5 Ray Reconstruction for better quality ray-traced graphics, alongside some interesting GeForce Now announcements).

To refresh your memory and recap — in a nutshell, as there's a lot more to it than this — Nvidia has billed DLSS 5 as a 'real-time neural rendering model' that polishes up a game's visuals using AI (without altering the game assets, and with the developer being given a decent level of control over the exact implementation).

What actually happened here was that on Friday, VideoCardz spotted that following NBA 2K27's early access going live, a DLSS 'neural rendering' DLL was discovered in the build. It didn't take long at all before modders (in the RenoDX community) fudged things to get this working in games ('SpeedLemur' being the principal modder who "made this happen").

So, in theory this is effectively our first glimpse at DLSS 5 in the wild, with those modders demonstrating what it looks like by using the third-person fling-stuff-about action-adventure Control – as seen on Reddit. However, do note that the tech is only working for application to character models for now (not the environment or objects).

Nuts and bolts

There have been some very interesting revelations from those experimenting with this apparent leak of DLSS 5 in terms of how it works, how it runs — and perhaps most telling is the broad reaction to this spillage, which I'll return to shortly.

First, the nitty-gritty. In this leaked form — so add a whole heap of seasoning, because nothing here is remotely official — DLSS 5 seems to have a small set of adjustable parameters, including seven different 'styles' and various sliders to play with intensity, tone and structure. We're told that neural rendering doesn't need much in the way of data to do its stuff, and works off much the same inputs as regular DLSS (basic upscaling – where all this started, remember?).

It's also for Nvidia RTX 5000 GPUs only (although that's still not fully confirmed, it was always likely to be the case anyway). The news on the performance front goes a bit awry, though, with the discovery that in Control the frame rate is cut in half by the AI polishing up the visuals. With an Nvidia RTX 5070 Ti running Control at 4K with ray tracing on high and DLSS (Quality), the frame rate was 71 fps and it dropped to 35 fps with neural rendering.

That is, of course, a nasty hit indeed. However, this is a piece of leaked code – in an unknown state as far as the stage of development is concerned (one clue suggests that the model may date back to the start of July) — extracted and fudged to work on an old game. Given all that, we can't take this as anything like an indication of what the finished version of DLSS 5 might be like performance-wise.

Although it's at least comforting that it works in a playable (enough) manner on a single RTX 5070 Ti graphics card, because if you recall, Nvidia's demos used not one, but two, RTX 5090 GPUs. Clearly a 50% frame rate drop won't fly, granted, but Nvidia wouldn't release this neural rendering trick if the hit was anything like that, because no gamer would use it in that state.

Change in tone

What we can really take from this leak and the modding efforts thereafter is quite limited, then. But what's clear is that as mentioned, there's been a marked turnaround in the commentary and debate on the image quality on offer with neural rendering.

As one Redditor observes: "Remarkable difference in tone of the reactions compared to when the [initial] Resident Evil footage was released."

Someone replying to that says: "People freaked out about what was obviously just a bad demo, and failed to see what was actually going on with the underlying tech.

"There were a few talking about how different tone mapping fixed the issue with minimal effort, or how it wasn't actually adding/changing details, and it turns out they were correct. People freaked out over nothing imo."

Another gamer notes: "This looks like legit magic. I was saying this when the reveal happened and people wanted to burn Nvidia down: the tech is amazing, the demo might be off because it's too heavy handed, but the tech itself is incredibly impressive."

And a further Redditor enthuses: "Insanely huge graphical update. There's no going back."

You get the idea, then, and it's quite an eye-opener how many comments there are along these lines. Quite a lot of gamers are impressed by this leaked spin on DLSS 5 — and I can see why, frankly, at least with the Control material presented here.

And remember, this is with DLSS 5 working on character models, with the facial tinkering and tweaking being one of the main causes of concern with Nvidia's initial demos. The likes of Flight Simulator addicts are positively drooling about what this might mean for the graphics of the landscapes beneath their virtual wings.

And there's a lot of excitement about how the tech might be applied to older games like Control to make them look a good deal better.

Skeptics remain

Naturally, not everyone is convinced. The detractors remain and are unswayed by the material which has popped up around this leak so far. They argue that the AI-enhanced Jesse (Control's protagonist) is still all too 'uncanny valley' in appearance, and that DLSS 5 is still very much a visual downgrade.

Others also worry that there aren't enough fine-grained controls for the developer in what we see here. Although again, we can't take this leak as evidence of what DLSS 5 will turn out to be eventually, whether that's the end result of the graphics themselves or the suite of dev controls under the hood.

The performance hit has been seized on, too — as noted, though we can't judge on that yet, not with an early leak — and some feel that while screenshots might look good, this doesn't hold up in motion.

Simply put: it is still too early to judge, and we need to grab large handfuls of seasoning with all this. But the general sentiment on DLSS 5 appears to be turning around in quite a sizable shift of opinion — and Nvidia may yet convert PC gamers to its AI-enhanced vision of the future (providing any performance hurdles are cleared, of course).

I'm currently holding the Samsung Galaxy Z Fold 8 in my hands, only putting it down now to type this and tell you that, as I see it, it's the nearest thing we have at the moment to an iPhone Ultra, the folding iPhone we expect to meet in person on September 9. And what I wonder is not just if the real iPhone Ultra will be good enough to glow up the foldable category so that it eventually rivals standard flagship phones, but whether if it can rise above the foldable bar set by Samsung and, to a lesser extent, its rivals.

The iPhone Ultra will arrive on the tiny island that is foldable phones amid a group of well-established competitors. Apple's slow walk into this category has allowed Samsung, Google, Huawei, Motorola, and Oppo to hone their skills in creating thin, light, and powerful folding phones, often featuring excellent screens, battery life, and cameras.

The Samsung Galaxy Z Fold 8 is not the best of them. Don't get me wrong, it's an excellent foldable phone, but it's not the thinnest, doesn't have the longest battery life, and has just two rear cameras. On the other hand, the passport shape (when folded) makes it compelling in a way that's hard to quantify and temporarily unique among current foldable options. When opened, it has a screen size that is nearly as large as those found on other foldables, but in an oddball 4:3 screen aspect ratio.

The design is intentional, with Samsung hoping to bridge the gap between the clamshell-style and fashion-forward Galaxy Z Flip and the far more powerful, productivity-first Samsung Galaxy Z Fold 8 Ultra.

Samsung is filling the gap in the hope of vastly expanding what remains a very tiny market. I know there are reports that Samsung is selling out the Z Fold 8, with one report claiming that it sold out and is now rushing out a million more units. Overall sales for the Z8 line (Fold, Flip, and Ultra) might be in the 4-to-5M unit range. That sounds pretty good, but a look at typical iPhone sales after launch quickly puts those numbers in perspective.

In 2024, iPhone 16 preorder sales reportedly reached 37 million units. Apple used to provide quarterly updates on iPhone unit sales, and they were, honestly, insane. In December of that year, the Cupertino tech giant sold more than 77 million units. The year ended with over 217 million iPhones sold.

What the industry believes, or at least hopes, is that the arrival of the iPhone Ultra cracks open the market in ways Samsung and its folding counterparts haven't yet. This notion is predicated on the assumption that the iPhone Ultra will be substantially better than Samsung's folding line.

Except it probably won't be. I fully expect it to be an exciting and excellent smartphone, but Samsung, more than any other company, has come close to perfecting the art of the fold. The Z Fold 8 Ultra unfolds flat, has an almost creaseless screen, and is as light and thin as a standard flagship, while offering twice the utility.

The folding iPhone : what we expect

Apple's folding iPhone is not even expected to match the Z Fold 8 Ultra in style and features. In fact, it's a fair bet it will be in a form factor almost exactly the same as the Z Fold 8. That's right, we all expect Apple to produce the same kind of Passport-style foldable, right down to the pair of rear cameras with no telephoto lens in sight. This shouldn't be too surprising since most believe Apple is sourcing its folding display from Samsung Display.

Among the rumors I've read about the 5.5-inch (folded), 7.8-inch (unfolded) iPhone Ultra is that it will be built with a combination of aluminum and titanium, just like the Z Fold 8. It will, at 4,5mm, be no thinner than the Z Fold 8 when unfolded. The chief design distinction might be the use of what's rumored to be called "liquid metal" to try and fully erase the crease (something the Z Fold 8 Ultra and Z Fold 8 have nearly done).

Apple might have an edge, though, where it counts: performance. Most expect the A20 chip paired with 12GB of RAM and Apple's own C2 cellular model. All of this could make this a faster foldable, even if the average user never notices.

Some believe that Apple doesn't need to make the world's best foldable; it simply needs to deliver it to a willing market. How many people have been holding off on buying a foldable because Apple doesn't offer one? That could be millions of people, and if that's the case, it might be the Apple cachet that grows the foldable phone market to a previously unimaginable scale.

Dennis Gabor was a Hungarian-British physicist who's widely credited with inventing holography, a photographic technique that can record and reconstruct three-dimensional images. This is an invention for which he received the Nobel Prize in Physics in 1971. But his legacy extends to his philosophy on the nature of innovation

Go your own way

Gabor wrote this iconic passage in his now-extremely rare book 'Inventing the Future', in which he sketched out an optimistic vision for the future.

At the core of his belief was that there was no need to any longer passively predict what may come to pass in the years ahead. Instead, the influence and strength of science and industry was such that mankind could forge its own future.

The comments also focused on that pivot point of the industrial revolution, before which the possibilities of the future were fairly limited, in his view. As each wave of innovation comes to pass, the possibilities continue to expand.

Charting a path

There's no understating the influence that Gabor had in modern science and industry. His comments were routinely quoted, and even adopted, by others. Chief among the figures that took this philosophy to heart was Alan Kay, described as one of the true luminaries of personal computing.

Kay, whose career spanned stints at Apple and HP among other industry titans, used this idea of inventing, not predicting, the future as a core maxim from 1971. He used the phrase to inspire his work in pioneering elements like the graphical user interface and laptops many years before they were realized.

Although Gabor was a very influential figure in the technology world, his predictions weren't always on the mark. In 1962, for instance, he said sending documents via telephone wires was possible, but only in principle. That's because the equipment required was so expensive that he didn't see a world in which they'd be cheap enough for the technology to become practical.

Since I started my monthly Blu-ray Bounty column in November 2024, I’ve tested more than 100 4K Blu-rays. In that time, as well as finding that a lot of highly anticipated discs — such as Fight Club — lived up to the hype, there have also been some pleasant surprises, including the reference-quality Speed Racer and the spectacularThe Super Mario Galaxy Movie. The latter two are now part of the selection of discs I use to test the best TVs and best soundbars.

While I haven’t managed to test every disc I’ve wanted to (and certainly not every 4K disc released), one has remained frustratingly elusive: Hard Boiled. One of John Woo’s most iconic films, it's a cornerstone of Hong Kong action cinema, and I'd been keen to get my hands on the 4K release.

After failing to snag the Limited Edition, released in March 2026 in the UK, and every subsequent restock, I was starting to wonder if a Hard Boiled was destined to stay on my testing wishlist. July saw the arrival of a standard edition — and it shot to the top of my July 2026 most anticipated 4K list. After yet more courier delays, Hard Boiled finally made it into our testing lab — so, at long last, I was able to give it a spin to see whether the 4K disc is worth all the hype.

Awesome Atmos audio

There are three soundtrack options on this disc: Cantonese Mono, Cantonese Dolby Atmos, and English Mono. I primarily used the Cantonese Dolby Atmos, to see how effective the upscaled audio track would be, although I was impressed by the clarity of the Cantonese Mono.

There're no shortage of action-packed scenes in Hard Boiled, in particular a number of gunfights, and the disc's Dolby Atmos soundtrack makes the most of them. During testing, I used a Samsung HW-Q990C 11.1.4 channel soundbar system, and the Atmos mix really came alive. As bullets tear through the room, ricocheting bullets can be heard bouncing off walls with impressive detail and precision. In the opening gunfight, as sprays of bullets from sub-machine guns fly, the trajectory of these bullets was precisely mapped to the rear channels, and as the camera moved around the room, the sound of the bullets moved with it.

While bullets generate impact, it’s the big explosions that deliver the meaty bass. While these explosions didn’t always resonate from the Q990C’s subwoofer (down to the older mix), there was still plenty of oomph to each one. The Atmos soundtrack also displays great balance and range. As Tequila and Ah Long are caught in a shootout with one another, the sound of bullet casings hitting the floor is audible in the loud gunfire, which is in turn well balanced with raised voices and the often-lively score.

4K picture wins again

Arrow Video’s 4K discs are among the strongest when it comes to restoration, and this remains true with Hard Boiled. Textures have been finely restored, with details such as skin taking on a remarkably natural, almost 3D quality. In some scenes, the movie looks like it could have been shot only a few years ago rather than in 1992. Close-up shots reveal an impressive level of refinement, too, from the texture of a gun to the tiny details of a bullet. Having recently visited Silver Salt Restoration, which occasionally works with Arrow Video, I know just how much hard work goes into restoring films to such a standard.

Elsewhere, this 4K disc showcases bold, vibrant colors that really pop off the screen. The neon signs of downtown Hong Kong have a gorgeous richness to them, while restaurant interiors benefit from punchy yet natural-looking hues – such as a vivid green door in the opening gunfight. And then there's the blood — of which there is plenty in Hard Boiled — that displays real vibrancy, but in some scenes, also appears deep, rich and convincingly realistic. It really highlights the disc’s color reproduction.

Throughout the movie, there are plenty of scenes that, displayed on the right TV, really highlight the 4K disc’s excellent contrast. In the dimly lit jazz club during the opening credits, the dim surroundings contrast well with the wall and stage lights. A later scene of a parade of motorcyclists escorting a car along the docks also show deep black tones, which balance nicely with the headlamps from the vehicles.

Finally, there are plenty of moments that can give a good TV a chance to shine, too. Throughout the movie, bright sparks fly from car collisions, showers of bullets and explosions, and on the LG G6 I used for testing, they really showed off the G6’s strong HDR highlights. The film grain has also been cleaned up. There’s still enough to satisfy cinephiles, but the clean-up job gives Hard Boiled a more contemporary look without stripping away any character.

My 4K Blu-ray love grows

I’ve written plenty about how much I love what 4K Blu-ray can do for more classic movies, with titles such as 1938’s Captain Blood, 1944’s Double Indemnity and 1950’s Sunset Boulevard from Hollywood’s Golden Age all given a new lease of life. Even "more recent" thrillers such as To Live and Die in L.A. — also from Arrow Video — have received stunning 4K makeovers, while remaining crucially faithful to the original, filmmaker-intended look.

Hard Boiled is just the latest in a series of titles I’ve tested that showcase just what 4K Blu-ray can do. It retains the movie’s iconic look, both of early-'90s Hong Kong cinema and John Woo’s distinctive style, while also giving the movie a new coat of paint. The superb new Atmos soundtrack adds another layer, bringing out a wealth of intricate detail and making the explosive action even more immersive. It's yet another fantastic 4K Blu-ray release that I’ll be adding to my now bursting 4K Blu-ray shelves.

Identity is the fundamental currency in any technology transaction so we need to treat nonhuman assets with the same care as we apply to people.

We’re outnumbered! Nonhuman identities (NHIs) outnumber human identities in the enterprise by a ratio of up to 50:1. This Non-Human Identity Management Group statistic from 2025 is likely already outdated, and in a year from now the ratio will be exponentially higher as rapid growth trends like AI/Machine Learning, the Internet of Things, digital assistants and the burgeoning API Economy continue to accelerate automation.

What are NHIs? Machines, processes, service principals and accounts, virtualized and other workloads, and applications that are granted digital credentials. Think, for an everyday example, of a helpdesk chatbot on your favorite website.

Managing NHIs is a challenge that CIOs, identity experts, and CISOs must accept. This in some ways is a re-run of what happened 40 years ago when identity access management was widely deployed. It’s just that now we are not only dealing with human beings. And I don’t know a single organization that has its collective head around the concept of who is, or should be, ‘the HR chief’ for NHIs.

Leaders today need to consider nonhuman identity in the way they consider human identity. When we hire, we onboard people by harnessing and securing the details we need to pay them and manage them, but we also work to imbue them with our organizational culture, systems, risks, security processes and so on.

We set and tweak permissions based on who needs access to what and when, depending on their roles, seniority, specific tasks they perform at a given time and so on. We need to have similar controls and contextual understanding of our nonhuman assets.

An urgent case where context is king

Without making that change CIOs and other leaders will have no foundational basis of understanding where risks are coming from or what measures they need to take. They need a central, visible platform to see what they have and what’s going on. Legacy IAM can’t offer that because it starts with a narrow focus on compliance and is a static model based on rigid criteria like six-month audits and the rote dispensing of permissions and privileges.

The old IAM is rooted in the early-21st-century age of Sarbanes-Oxley, reactions to governance scandals and corporate malfeasance. Moreover, it ignores what is happening in the wider digital world: namely, the explosion of identity types that are crying out for holistic oversight.

Control of the human and nonhuman estate means ensuring there is insight into nonhuman assets and what they're doing.. context is king. So look at NHI constructs that can touch anything in the digital ecosystem and the underlying connectivity to apps and, most importantly, data. CIOs need to be able to recognize the truth of a contemporary mantra: identity is the fundamental currency in any technology transaction.

Everything must be covered: bots; service principals; IoT endpoints; abandoned trial workloads; autonomous agents and agents that speak to other agents. Start by asking an overarching question: do these NHIs leverage a human identity when they perform tasks on behalf of that person, and is that appropriate?

Then it is possible to drill into practical areas like guarding against credential misuse across agents, improving digital and identity hygiene by spotting risks relating to non-expiring tokens, and analyzing API access levels on a just-in-time basis.

ISPM (Identity Security Posture Management) is useful here because organizations need to have a handle on what they have, and must bring identity constructs together. However, this is not a ‘one and done’ discovery process so consider it as a CI/CD journey of continuous improvement.

In the AI era, holistic technology inventory can't be CMDB-based but must be dynamic and based on CI/CD pipeline control frameworks that let us dynamically view and slice and dice data and assets by controlling who is using what and where.

Think of it as an opportunity…

It’s a complex task and we are all busy. But we need to pause and think about how this inflection point can be parlayed into a positive. This is an opportunity for the security and identity teams to start asking ourselves about the constructs we can apply to nonhuman identities.

They should be asking what needs to be added to legacy IAM frameworks to be successful. And usually the answer will be to replace them with systems that are coded for the current age, not the previous one.

By providing these new controls the identity team will become respected and sought out. Identity leaders will gain kudos by their ability to attribute risk ratings based on behavioral analytics and business function.

By demonstrating insights into where data sits, where it moves to and from, and how identity supports that, identity leaders can deliver qualitative and quantitative process analysis. In some cases of outperforming best practices and execution, they will even help to refine business process re-engineering through a platform that informs prioritization and budget management.

We've featured the best IT management tool.

This article was produced as part of TechRadar Pro Perspectives, our channel to feature the best and brightest minds in the technology industry today.

The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/pro/perspectives-how-to-submit

In the past eighteen months, our teams have moved from debating whether to use AI to debating how fast we can deploy it. The harder question is how to let teams move quickly enough to capture the value of AI without allowing the company’s risk profile to expand faster than its ability to govern it.

That tension is familiar to technology and security leaders because AI creates two mandates that can appear to compete with each other. The technology side of the organization wants experimentation, access, speed, and a path to real productivity gains, while the security side needs control, accountability, data boundaries, and confidence that new workflows will not introduce avoidable exposure.

Both instincts are correct, which is why companies get into trouble when they treat AI as either a pure innovation project or a pure security problem. It is an operating model change, and the organizations that handle it well will be the ones that build just enough structure and hardened tools to let teams move with confidence rather than forcing them to choose between speed and control.

Start with the work, not the tool

Many companies begin by treating AI adoption like a standard software rollout. They approve a vendor, distribute licenses, publish a few guidelines, and assume usage will naturally become transformative. That approach can create activity, but it rarely creates durable operational change.

Real adoption starts when leaders understand how work actually gets done. A finance team, product team, marketing team, support team, and engineering team will not use AI tools in the same way because each group has different knowledge requirements, data sources, risk thresholds, and experience. Each of these will require the development of AI skills and tools.

Leaders should begin by asking what each function is trying to accomplish, what knowledge it needs to make better decisions, what skills are required to use AI responsibly, and what tools or data sources are necessary to produce a reliable result.

When AI is mapped to those capabilities, it becomes part of how the organization operates; when it is layered on top of disconnected processes, it tends to create more output without necessarily creating better outcomes.

Treat data access as a risk design problem

Data access is one of the clearest places where AI changes the operating model. To make AI genuinely useful, teams often need access to information they did not previously use directly.

A marketing team may need product documentation and customer insights, a finance team may need structured operational data, and a support team may need internal knowledge, historical context, and the ability to understand patterns across systems.

The instinct to open access is understandable because AI becomes more valuable when it has more context. The instinct to restrict access is also understandable because broader access can create privacy, compliance, and security concerns. The answer is not to choose one instinct over the other but to create a more deliberate model for deciding which data can be used, by whom, and for what purpose.

In one enterprise rollout, for example, the practical answer was not to give every team access to everything or to keep AI locked inside a technical function. It was to separate lower-risk operational data from more sensitive information, then give teams enough access to work differently while limiting the potential damage if a workflow behaved unexpectedly.

One useful way to think about this is the blast radius of data. Not every dataset carries the same level of risk, and not every AI use case deserves the same level of restriction. Some information can be made more accessible because the potential damage is limited if something goes wrong, while other information, especially personally identifiable information or sensitive customer data, requires much tighter controls.

This approach allows teams to experiment where the risk is lower while preserving stronger governance where the business truly needs it. The goal is not to make security lighter, but to make it more precise so the company can move faster without losing control of the environments, data, and workflows that matter most.

Build verification into the workflow

Verification is where many AI strategies either become scalable or begin to stall. AI systems can produce work that looks polished but is incomplete, inaccurate, off-brand, or noncompliant, and while human review can absorb some of that risk in the early stages, it does not scale well once AI becomes embedded in daily operations.

I have seen this most clearly in technical teams, where AI can accelerate software development only if there is a reliable way to evaluate the quality, security, and accuracy of what gets produced. Without that verification layer, teams may feel faster in the moment while quietly creating more review burden, more rework, and more risk downstream.

Companies need verification patterns that are designed into the workflow itself. That can include automated checks, data quality controls, approval paths, logging, observability, and clear escalation rules for outputs that should not be trusted without further review.

Leaders shouldn't treat these mechanisms as bureaucracy added after the fact. They are what allow AI to move from individual productivity aid to enterprise capability, because teams can only scale adoption when they have a repeatable way to understand whether the work being produced is accurate, appropriate, and safe to use.

Make AI adoption a leadership responsibility

AI adoption cannot be delegated to a single innovation team or AI officer while the rest of the executive team watches from a distance. Every function leader needs to understand how AI changes the work their team performs, how employees will collaborate with agents and models, and where the risks are likely to appear.

That does not mean every executive needs to become a machine learning specialist, but it does mean leaders need enough fluency to guide decisions, set expectations, and model the behavior they ask of their teams. Employees are unlikely to change the way they work if their leaders treat AI as something other people are supposed to adopt.

The cultural framing also matters because if employees hear AI adoption as a euphemism for job elimination, they will protect the current version of their role rather than explore what the next version could become. Leaders should be honest that roles will change, some workflows will disappear, and new responsibilities will emerge, while also making clear that the near-term goal is to help people do higher-quality work with better leverage.

In many cases, the shift is from doing every task manually to orchestrating systems that help do the work. That requires a willingness to work through ambiguity, which becomes more important as AI becomes more capable.

The companies that succeed with AI will not be the ones that put innovation ahead of security and they will not be the ones that wait until every risk can be eliminated in advance. They will be the ones that build enough governance, verification, and data discipline to let teams move with confidence, because within enterprise AI, control is what makes speed sustainable.

We've featured the best endpoint protection software.

This article was produced as part of TechRadar Pro Perspectives, our channel to feature the best and brightest minds in the technology industry today.

The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/pro/perspectives-how-to-submit

If the past few months are anything to go by, extreme heat is becoming part and parcel of the British summertime. As temperatures continue to rise, employers are under growing pressure to make sure workplaces can cope with the hotter conditions.

The challenge is understanding the smartest way to adapt, because the cost of leaving it to guesswork and getting it wrong is significant.

The UK’s June 2026 heatwave is estimated to have resulted in 24 million lost working hours and a cost to the economy of £1.15 billion in lost productivity. Those figures show the direct impact workplace conditions can have on business performance.

At the same time, hybrid working adds another layer of complexity, having fundamentally changed how offices operate. Attendance is far less predictable than it once was, so it’s harder to know where and when things like cooling, ventilation and workspace capacity are needed most.

Aside from the summer heat, organizations need workplaces that perform well in every season and adapt to constantly changing occupancy and environmental conditions. They need to understand how people move through the office. Things like which desks they check into, which meeting rooms fill up, where teams collaborate and which spaces are regularly ignored.

Doing so requires a real-time understanding of how buildings are being used. This is where sensor technology and workplace usage data comes in, turning assumptions into evidence and giving organizations the insight to make smarter workplace decisions.

Building-wide averages can hide local problems

Many offices are still managed around a relatively fixed idea of how a building should operate. Cooling and ventilation systems may follow predetermined schedules, while individual spaces are managed according to their maximum capacity, rather than how they’re being used in the moment. But we know modern workplaces don’t behave that predictably.

One day a floor may be largely empty, while another is full. The following day, those patterns may be completely different. A busy Tuesday might see meeting rooms become warm and stuffy after hours of back-to-back collaboration, while the open-plan workspace next door remains comfortable. Likewise, areas exposed to direct sunlight will naturally heat up much faster than others.

A single temperature reading taken in one location simply can’t capture those differences. Yet building-wide decisions are often made using exactly that kind of limited information, resulting in energy being used where it isn’t needed while genuine comfort issues remain unresolved.

That’s why employers need a more complete view of workplace performance. Environmental sensors can monitor temperature, humidity, carbon dioxide and air quality across different areas, while occupancy data can show which spaces are being used and how it changes throughout the day.

Often, organizations assume they need to monitor employees to understand workplace performance, but that’s not the case. Built with privacy in mind, systems can combine thermal sensing, ambient temperature, CO2 levels and anonymous spatial tracking to show how an area is performing without needing to identify the people using it.

Bring those data points together, facilities teams can identify real issues earlier, understand what’s causing them and prioritize action that have real impact for employees and their productivity.

Making the data useful

Collecting workplace data is only the beginning. The real value lies in how organizations use that information to create better experiences for employees while operating buildings more efficiently. It’s easy to focus on dashboards and sensors here, but the real goal is to give facilities teams information that’s helpful and they can actually use.

Workplace intelligence isn’t just about responding more quickly when something goes wrong. It’s about anticipating issues before they begin affecting the people using the space.

For example, rising carbon dioxide levels may indicate that a meeting room needs better ventilation, while recurring temperature spikes could point to equipment use, afternoon sunlight or an imbalance between occupancy and cooling. This allows facilities teams to pinpoint the root cause of problems and take action before they begin to impact employees.

By directing resources to the areas that genuinely need them, employers improve employee comfort while reducing unnecessary energy use and operating costs. This more targeted approach also supports wider sustainability goals.

Avoiding unnecessary cooling across half-empty floors reduces operating expenditure and helps organizations make progress against Scope 2 emissions targets, without compromising productivity or indoor air quality.

Designing workplaces around how people really work

Perhaps the greatest value of workplace data lies in what it reveals over time.

Patterns begin to emerge, so organizations build a much richer understanding of how their workplaces are being used.

Recurring patterns can reveal which spaces are consistently underused, which place the most demand on building systems and whether certain areas should be opened, consolidated or redesigned altogether. They can also inform more accurate and efficient cooling schedules, room-booking policies and decisions about where particular teams or activities are best placed.

Check-in and booking data add another layer to that. An organization may discover that employees regularly choose a particular bank of desks, that small meeting rooms are constantly oversubscribed or that a purpose-built collaboration area is rarely used. Those behaviors tell facilities teams something that a temperature reading alone never could.

That said, data is most useful when it is considered alongside employee feedback.

For organizations managing several offices and thousands of employees, speaking to everyone individually isn’t always realistic. Workplace data can help identify where deeper conversations are needed, while surveys, feedback tools and direct employee input provide the context behind the numbers.

Workplace intelligence can highlight what’s happening across a building, but employees provide the context that explains why it matters and how conditions affect their ability to collaborate and concentrate.

Bringing these perspectives together creates a much more complete understanding of workplace performance.

Ultimately, it’s about creating offices that continuously adapt to changing conditions to keep employees comfortable and productive, rather than relying on static assumptions that quickly become outdated.

Preparing offices for a warmer future

It's not only the weather that has become harder to predict. Working patterns are changing just as quickly, and static approaches to workplace management are increasingly out of step with both.

Organizations need to understand both how a building is designed to perform and crucially how it’s being used from one day to the next. That visibility enables facilities teams to respond earlier, use energy more intelligently and create environments that remain comfortable and flexible as conditions change.

The benefit is a workplace that simply works better for the people using it. Somewhere people can focus when they need to and trust that the environment will support the work they have come in to do.

Put simply, when a workplace works better for its people, they are far more likely to want to use it.

We've featured the best business computer.

This article was produced as part of TechRadar Pro Perspectives, our channel to feature the best and brightest minds in the technology industry today.

The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/pro/perspectives-how-to-submit

The paperless office is one of the clearest examples of the impact that digital transformation has had on workflows. This includes industries, like law and finance, where physical records of compliance documents, market-sensitive data, and customer information must be retained.

Mishandling electronic and physical data can lead to heavy regulatory penalties, so companies take care to encrypt data in transit, enforce MFA, segment their networks, and sharpen endpoint detection.

These measures build a solid foundation of security. But what good is that wall when the weakest link sits in the corner of the office, quietly printing?

Regulated industries have a print security problem

Most offices treat printers as passive output devices, disconnected from any threat. That assumption is out of date. HP found 57% of IT decision-makers rate print security as a low priority in their cybersecurity strategy, and 45% aren't confident their print environment meets compliance standards.

Whilst most print-related losses differ in complexity, they can be easily traced. Data leaks expose unencrypted hard drives and unpatched firmware, while print history can reveal when documents have been left sitting in the output tray.

Modern printers are networked endpoints with access to confidential cloud environments and increasingly, they're managed by third-party MPS vendors, widening the potential attack surface.

Managed Print Services are useful for fixing hardware gaps, but they can introduce new risks. When print jobs are routed through third-party cloud servers without end-to-end encryption, files in transit become exposed. Granting an external MPS vendor administrative network access creates backdoor entry where even if only the vendor is compromised, attackers can easily pivot into the corporate network.

Singapore saw this play out in 2025, when a ransomware attack on a printing vendor for Bank of China's Singapore branch exposed 11,200 customer names and account details. The incident is a clear example of how a print vendor can easily become the weakest link in an otherwise solid network.

Hybrid work adds another layer of complexity and few companies have clear governance for print behavior once it leaves the office. Office workers may have enterprise-grade tools, but remote workers don’t. Employees printing at home may be using unmanaged devices and unsecured Wi-Fi, offering little visibility into what's being printed or where it ends up.

The compliance exposure under FCA rules

When two-thirds of knowledge workers assume network printers are secure by default, the likelihood of a printer-related breach goes from a matter of if, to when.

For UK financial services firms, this creates an operational and regulatory risk. The FCA has no statutory cap on fines, and its expectations around data governance extends to how firms handle information physically, not only digitally.

Regulators won’t distinguish between a breach caused by a compromised server and one caused by an unattended output tray; the obligation to protect customer data is the same either way. Conversely, a firm can encrypt data in transit, enforce MFA, and segment its network, and still face exposure if a printer or an MPS vendor becomes the point of failure.

Security starts at the print queue

Even as workflows digitize, printers aren't going anywhere. Regulatory demands still call for hard copies and inked signatures, putting print infrastructure at the heart of broader conversations about automation and intelligent document processing.

Every print, scan, or routing job sits in between physical and digital workflows, making them a powerful lever for end-to-end process improvement. Cybercriminals know this too and lapses in printer security become an open door to confidential data.

Content-aware and cloud-based print management platforms close it. By doing away with physical print servers, they allow businesses to centrally manage printers, automate driver installs, and enforce secure "pull printing," where documents release only once a user authenticates at the device.

Because many cloud based print platforms run in Azure, the same security governing email and Teams can extend to print. Documents are checked against content-aware rules that determine whether they're printable or reportable, with flagged jobs blocked or escalated automatically.

Businesses can also go cloud-agnostic through S3-compatible integrations to avoid vendor lock-in, keeping secure releases separate from the product, so nothing installs on the printer itself and everything is managed remotely.

The last unsecured endpoint

Ultimately, security is a board-level concern, and printers need to be treated as the endpoints they are. That means extending endpoint detection and firmware patching to print fleets and extending governance to remote printing, even if that means restricting what can be printed at home altogether.

This is where visibility matters most. In third-party risk management, paper trails are supposed to tell a coherent story but rarely do. Evidence of oversight is too often scattered across email threads, physical documents, and local files. Vetting MPS vendors is how businesses close that gap, but the work starts with treating print infrastructure as seriously as any other endpoint on the network.

We've featured the best firewall software.

This article was produced as part of TechRadar Pro Perspectives, our channel to feature the best and brightest minds in the technology industry today.

The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/pro/perspectives-how-to-submit

Of the expected $2.5 trillion that corporates are due to spend on AI this year, over $1 trillion is set to be allocated to IT services and dedicated software. That is a huge chunk of change for a technology that many are still grappling with and trying to understand how it fits with their business, team, and procedures.

A few years since the ChatGPT-enabled revolution started, we’re beginning to see the different approaches and challenges businesses are facing when implementing AI.

Ford's recent decision to bring back experienced engineers after AI systems failed to catch manufacturing issues offers an important lesson for every business investing in AI. At the same time, new research from Ramp and Revelio Labs, covering almost 22,000 US companies, found that organizations investing most heavily in AI are actually hiring more people, not fewer.

Together, these stories tell us something important. Everyone assumed AI would reduce reliance on people, but instead we’re seeing companies discover how dependent AI is on experienced people. AI may be changing how work gets done, but this alone isn't enough for growth and success.

AI is accelerating work, but experience still matters

Earlier this year, Ford re-hired 350 veteran engineers after AI and automated systems failed to deliver quality work, demonstrating that employees are critical to a successful AI rollout.

It’s also interesting to see this reflected in Ramp and Revelio Labs’ research that organizations investing in AI are hiring more people. Instead of neglecting hiring and skills development, companies increasingly need the right team members to help refine and improve sophisticated AI so that it performs better in the areas they need.

Success with AI still depends on understanding the skills people need to perform, at an individual, team and organizational level, then investing in development that reflects the realities of their work. But before organizations can build those skills, they need to know where expertise already exists. Too many businesses are sitting on critical knowledge without any clear view of who has it or where the gaps are.

One way to achieve this is building a skills ontology, a living record of the skills that exist across the organization.

Rather than relying on assumptions, businesses can identify where they overindex or underindex on capabilities, understand which teams have specialist knowledge and spot gaps before they become business problems. It also creates the foundation for workforce planning, internal mobility and more effective AI deployment.

The most valuable skills inside your organization

Now you have an effective skills ontology, you have an overview of what is going on in your organization. This helps you identify the subject matter experts (SMEs) that exist across a business, but whose knowledge has often never been formally captured - such as that mythical AI ‘power user’. Take Ford, which has said it is using its rehired employees specifically to train younger staff and reprogram AI tools.

Once you’ve identified where expertise exists, AI can help maintain that picture and keep it up to date as roles evolve. But understanding where the expertise exists is only the beginning.

Organizations need to validate that expertise, connecting people to targeted development opportunities and making it easier for SMEs to share what they know with others. Importantly, this learning should be embedded in work as much as possible rather than separated from it, so it improves performance where it matters most.

AI makes knowledge easier to organize and distribute, but it doesn’t create the judgement that comes from experience. It can surface patterns and recommend learning pathways, but it can’t explain why a process works, when to challenge it, or how to adapt it to the realities of your organization.

That’s why experienced team members remain essential. They contribute the context that enables AI, and everyone else, to use that knowledge effectively.

Turn expertise into a competitive advantage

At a time when many in areas like HR and Learning and Development (L&D) felt that AI would diminish their role, it's clear that this expertise is only becoming more valuable.

L&D teams have the opportunity to operate more strategically by owning the upskilling and reskilling agenda, turning their and team members’ expertise across the business into a competitive advantage, and putting the processes in place to create agile workforces that collaborate effectively with AI.

AI has put organizational expertise firmly in the spotlight, whilst also highlighting that too many businesses are sitting on a wealth of knowledge that exists only in the heads of experienced employees. Not only do they know and understand more about a business and its processes, but they can also provide judgement - a scarce resource that AI cannot replicate.

Getting this information out and deploying it across the business will create a workforce that becomes smarter every time someone learns, rather than losing knowledge when someone leaves, leading to fewer repeated mistakes, better quality and stronger internal mobility.

To get the greatest return from their employees and from AI, the top organizations need to ensure they are investing in identifying, sharing and scaling the expertise they already have instead of simply buying more software.

We've featured the best online learning platform.

This article was produced as part of TechRadar Pro Perspectives, our channel to feature the best and brightest minds in the technology industry today.

The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/pro/perspectives-how-to-submit

Artificial intelligence may dominate today's headlines, but the most productive conversations I'm having across the technology ecosystem rarely begin or end with AI.

More often than not, they start with a business challenge.

Business challenges are driving the AI conversation

How do we modernize our infrastructure without increasing complexity? How do we strengthen our security posture while enabling innovation? How do we prepare our organization for AI without rebuilding everything from the ground up?

Those questions quickly lead to a realization: today's technology decisions are interconnected.

An AI initiative depends on cloud readiness. Cloud strategy influences cybersecurity. Identity, governance and data management underpin all of it. On paper, they look like separate line items. In practice, they're one system. If you pull on cloud, security moves with it; if you pull on AI, data and governance move too.

That's a significant shift, not only for organizations investing in technology, but for the partners helping them navigate it. Organizations are moving beyond asking whether AI matters. They're focused on how to implement it successfully.

That shift is happening across the industry.

Our fourth annual Direction of Technology Report, which surveyed more than 1,400 IT solution providers across 40 countries, found that nearly 75% of partners say AI is essential to their future. Yet the same research shows the conversation has become much broader than AI itself. Partners increasingly see services, specialization, cybersecurity and automation as the capabilities that will differentiate them in the years ahead.

To me, that's one of the most important signals in today's market, because the conversation isn't simply about adopting the next technology. It's about building an environment where technologies work together to solve real business challenges.

From experimentation to implementation

AI adoption is accelerating, but organizations quickly realize that successful AI initiatives depend on much more than selecting the right model or platform. They depend on secure infrastructure, trusted data, governance, integration and the expertise to bring those pieces together.

The market data tells a similar story.

Global AI spending is projected to exceed $2 trillion in 2026. Global Public Cloud spending is expected to surpass $1 trillion, while cybersecurity spending is projected to continue its rapid growth. These markets aren't growing in isolation but expanding together because organizations recognize that one investment reinforces the others.

That's why I believe the conversation is maturing.

The question is no longer, "Where do we start with AI?"

It's, "How do we build the right foundation to make AI successful?"

That requires a different approach from everyone in the technology ecosystem.

Customers aren't asking for another AI demonstration, but they’re asking questions like:

• How does AI fit into the technology investments we've already made?
• How will it integrate with our existing cloud environment?
• What does it mean for our security posture and governance?
• How can it improve productivity without creating unnecessary complexity?

Those are implementation questions, and I believe they'll define the next phase of AI adoption.

Technology will keep evolving, and AI will evolve fastest of all. What gives me confidence is not the pace of that innovation but the way the industry is organizing itself around the harder work of implementation.

No single vendor, provider or advisor holds every answer. Environments have become too interconnected for that, and the organizations making genuine progress are the ones drawing on combined expertise, weighing infrastructure, security, data and services together rather than in sequence.

The definition of success is changing with it. What matters is no longer how many technologies an organization can reach, but how well those technologies operate as one system in service of a business outcome. That, far more than any single breakthrough, will determine who ultimately gets value from AI tools.

We've featured the best AI website builder.

This article was produced as part of TechRadar Pro Perspectives, our channel to feature the best and brightest minds in the technology industry today.

The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/pro/perspectives-how-to-submit

For years, retailers have competed on price, product range and customer experience. Those things still matter, but they are no longer enough on their own. Increasingly, success depends on whether customers can discover a retailer in the first place.

That is becoming more difficult. AI-powered search, changing platform algorithms and rising advertising costs are reshaping how people shop online. Customers no longer follow a straightforward journey from search engine to website. Instead, they move between marketplaces, social media, review platforms and increasingly AI-generated recommendations before deciding where to buy.

More than a third (35.1%) of UK businesses say reaching customers online has become harder over the past two years. At the same time, while 81.3% have a website, only 12.4% accurately track how website visitors become leads or customers. Almost half (45%) do not track this at all.

For retailers, these trends point to a bigger challenge. As customer discovery becomes more fragmented, businesses are becoming increasingly dependent on digital channels they do not control.

Retailers are relying on platforms they do not own

Search engines, marketplaces and social media have transformed retail over the past decade. They have enabled businesses of every size to reach customers far beyond their local markets and compete with larger brands. However, they have also created a new dependency.

Retailers invest significant time and budget building visibility on third-party platforms, yet those platforms are constantly evolving. Search rankings change, marketplaces adjust how products are displayed, advertising becomes more expensive and AI is beginning to influence which retailers customers see before they even click on a website.

These changes are outside a retailer's control, but they can have a direct impact on traffic, sales and customer acquisition. It does not mean businesses should move away from these platforms as they still remain an essential part of any digital strategy. The challenge is ensuring they do not become the only route to customers.

AI is making customer discovery more complex

AI is adding another layer to an already crowded digital landscape - instead of browsing multiple websites themselves, customers are increasingly asking AI tools to recommend products, compare options or summarize reviews.

These systems draw on information from across the web, meaning businesses have less visibility into how recommendations are made and why one retailer appears ahead of another. For retailers, this creates uncertainty.

It is becoming harder to understand where customers first encounter a brand, which touchpoints influence purchasing decisions and how different digital channels work together. Rather than simplifying customer acquisition, AI is making it more dynamic. Businesses need to adapt by strengthening the parts of their digital presence they can influence directly.

Your website remains the strongest foundation

As new discovery channels emerge, it is easy to focus attention on the latest platform or technology. The more important investment is often the one retailers already own. A website remains the only digital channel where retailers control the customer experience from beginning to end. They decide how products are presented, how information is structured and how customers move through the buying journey.

The information published on a website becomes very important as it also helps shape how businesses are understood elsewhere online. Accurate product descriptions, structured content, customer reviews and up-to-date information all contribute to a stronger digital presence.

While retailers cannot influence every algorithm or AI model, they can ensure the information these systems rely on is accurate, consistent and trustworthy.

Visibility without measurement creates blind spots

Building visibility is only one part of the challenge. Understanding what happens once customers arrive is equally important. Yet our research suggests many businesses are still operating without a clear picture of how their digital channels perform.

Only 12.4% of UK businesses accurately measure how website visitors become leads or customers, while almost half do not measure this journey at all. Without that visibility, retailers cannot confidently identify which marketing activity is generating sales, where customers are leaving the buying journey or which investments deserve greater attention.

As customer journeys become more fragmented, relying on assumptions becomes increasingly risky. Retailers that connect website analytics, ecommerce platforms and customer data are far better placed to understand how people discover their business and what ultimately drives revenue.

Building resilience for the future

Retailers will continue to invest in search, social media, marketplaces and AI-driven discovery because that is where customers spend their time. However, long-term success will depend on balancing those opportunities with investments in the digital assets they own.

That means maintaining a high-quality website, improving product information, understanding customer behavior and measuring performance across every stage of the buying journey. These foundations not only improve the customer experience but also help retailers respond more effectively as technology continues to evolve.

AI will undoubtedly change how people discover products online, but it will not change one fundamental principle. Retailers that build direct relationships with customers and invest in the channels they control will always be in a stronger position than those relying entirely on platforms that can change overnight.

We've featured the best productivity tool.

This article was produced as part of TechRadar Pro Perspectives, our channel to feature the best and brightest minds in the technology industry today.

The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/pro/perspectives-how-to-submit

When was the last time you reviewed your cyber security standards? Let’s be clear: keeping a company secure is a constant task.

Cyber breaches are on the rise; according to the latest UK Government Cyber Security Breaches Survey, 4 in 10 businesses were compromised in the last 12 months. It’s important to remember businesses of all types and sizes are at risk – where there are gaps in security, attackers will take advantage.

Against this backdrop, plus high-profile cyber-attacks on companies including M&S and JLR, the government has reviewed its Cyber Essentials framework.

Its latest overhaul is raising the bar for organizations of all sizes.

Designed to provide the basic controls to protect businesses and their staff, if taken seriously, Cyber Essentials can make all the difference. Yet many businesses are still struggling to meet even baseline security standards.

Common issues such as a lack of budget, limited resources and a focus on other priorities are all factors holding organizations back – and they all make a major IT breach more likely.

Historically treated as a once-a-year compliance exercise, the latest ‘Denzel’ Cyber Essentials framework signals a clear and overdue shift towards implementing continuous cyber resilience. This means tougher standards which could catch businesses out.

Here are the key changes and how to approach them, to reduce the risk of a breach.

Implement a robust patching regime

Under the new framework, expectations have been tightened around patch management – the process of fixing a security risk or software error by installing updates – and vulnerability management.

Previously businesses could demonstrate compliance with patching requirements though documented processes and periodic updates.

In reality, teams find it hard to build in regular patching routines which work without affecting live services, due to downtime and continual testing. And that’s without taking into consideration all the third-party patches that need to be applied, as well as the critical operating system ones.

There is evidence that not addressing application issues and vulnerabilities is leading to more breaches. The 2026 Verzion Data Breach Investigations Report shows system compromise is now up to 61%, the highest it’s been over the last three years.

Timelines for applying critical patches have been made stricter, with a 14-day window to comply. Critical patches must be deployed consistently within this period, and businesses must be able to evidence their patching regime is working.

Companies will be afforded two chances to prove patches have been successfully deployed using a process called “double sampling”. If the first round of evidence has critical issues identified, a second chance will be given to fix it and prove it has worked – if it hasn’t, businesses will fail their certification.

IT teams should review systems now to determine if everything is being patched within the 14-day window. If not, they will need to work closely with operational teams to build up to the requirements in time for the next assessment.

Deploy Multi Factor Authentication (MFA) across the board

MFA remains one of the most effective controls against account compromise. The issue for many organizations is not a lack of MFA capability, but inconsistent deployment.

Businesses are likely to have MFA enabled for remote access such as Virtual Private Networks (VPNs) and Microsoft 365 services – but admin interfaces, cloud platforms and third-party services often do not follow the same practice, leaving companies exposed. Threat actors know this and use these weaknesses to try to gain access.

Under the new framework, if you have a cloud system which supports MFA then it must be enabled for all users, irrespective of whether it’s a free, included or paid-for option. Without MFA enabled, it’s an automatic certification failure.

This begs the question: do you know if MFA is enabled on all your cloud services? If not, how long will it take to turn this around?

Cloud services must be accounted for

Under previous versions of the framework, cloud services could be excluded from the assessment as it was argued they sat outside of the scope. This argument is no longer viable. Organizations need to provide a clear, well-defined document with evidence to support proper segregation of all IT Systems, otherwise they will automatically be viewed as within scope.

This means businesses must account for all services, regardless of where they are hosted. Given the flexibility of cloud solutions, this can quickly become a complex issue to gain appropriate visibility of all systems. If you are processing or storing data in a cloud service, this will be included. Typically, these are CRM platforms, HR software and financial systems, project management solutions, plus many more.

As it can take time to discover what’s in use and in scope, this is another area where organizations need to leave plenty of time to ensure all cloud services are covered.

Replace legacy hardware and update operating systems

A major change, and possibly the greatest headache, is the requirement to replace ageing hardware and software.

Recent retirement of the Windows 10 operating system is one area where businesses are under pressure to replace all their hardware to remain compliant. It doesn’t stop there – ageing IT infrastructure, along with major vendor services and solutions, all need to be replaced at some point. If organizations don’t plan ahead, they could be left with a mounting problem and not enough time to deal with it.

Once a vendor stops supporting a platform, the risk of a breach increases until this is addressed. Security patches will no longer be made available and given the time these systems have been out in the wild, it’s only a matter of time before another vulnerability is discovered and exploited.

In the short term, businesses should ensure they have a clear view of their assets in scope of Cyber Essentials and check vendor support dates (and when these run out) to avoid any hidden surprises at the last minute. In the long-term, a robust strategy for replacing all hardware as it is approaching end of life will save difficulty further down the line.

Don’t delay on a gap analysis

Core areas in the Cyber Essentials framework have been updated to reflect the fast-changing security landscape. It’s not all bad news, as businesses have time to turn things around. Organisations should read carefully through the new standards and consider how they apply to their own company.

Do not treat Cyber Essentials as a tick-box exercise and wait until a month before your next assessment. Do a gap analysis now and deal with the findings.

At the end of the day, it is absolutely worth the effort.

The best internet security suites for PCs, Macs and mobile devices, reviewed by the experts.

This article was produced as part of TechRadar Pro Perspectives, our channel to feature the best and brightest minds in the technology industry today.

The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/pro/perspectives-how-to-submit

I’ve been reviewing ereaders and epaper writing tablets for nearly a decade now and, in that time, I don’t think I’ve been more excited about an E Ink device as much as Amazon’s Kindle Scribe Colorsoft.

When it was announced in October 2025, I was keen to see how well Amazon had implemented the E Ink Kaleido 3 screen technology on the 11-inch note taker, especially since I found the 7-inch Kindle Colorsoft ereader to be one of the better-optimized displays in this category of tablets.

I had to wait over six months after it was initially available to buy in the US — living in Australia meant I could finally lay my hands on it in June this year — but I happened to try the 3rd-generation Kindle Scribe with frontlight first. Following that up with the Colorsoft version was probably a bad idea because I think the 2025 Kindle Scribe is arguably the best Kindle Amazon has made to date: it’s fast, the implementation of the monochrome E Ink Carta 1300 screen is fantastic and the battery life remains just as impressive as on any other Kindle.

The Kindle Scribe Colorsoft, in comparison, lacks luster. For the most part, it’s not really Amazon’s fault that its monochrome sibling outshines it — it’s the limitations of color E Ink display tech after all — but the one thing that really stands out to me is very much under Amazon's control.

Why so blue, Amazon?

I love the new wallpaper artwork on the 2025 release of the Scribes. They're beautiful, different and, after seeing them first on the 3rd-generation Kindle Scribe in black and white, I was keen to appreciate them in color.

Strangely, the dominant color you see on the lockscreen artwork on the Kindle Scribe Colorsoft is blue. From the start screen, showing the silhouette of the boy reading under the tree, to the red-and-white balloon hovering over a lake and forest, every default artwork just looks… blue.

Now, I know some of this is due to the screen itself — the color filter array (CFA) layered over the screen tends to shift hues towards a grey-blue. But if Boox can ensure the other colors pop in the default wallpapers on its large-screen color devices, so can Amazon.

I thought I’d get used to it but, two months on, and I’m left feeling… well, blue every time I wake up the Kindle Scribe Colorsoft. I don’t typically like displaying book covers on the lockscreen, but that’s what I’ve reverted to recently just to avoid the dominance of that one color.

It’s there in black and white

OK, I’ll admit that I sound like I’m nitpicking with the wallpapers, but it highlights one big issue with color E Ink displays — they’re darker than the monochrome E Ink Carta screens and, unless you bump up frontlight brightness, the colors don’t really pop. You can see that with not just the wallpapers, but with book covers and comics too on the Kindle Scribe Colorsoft.

This is because of the RGB layer integrated into the display and it affects every single color ereader and digital notebook, not just the Colorsofts. That means the ‘white’ background isn’t as white as it would be on a similar device using the E Ink Carta 1300 screen (like the 2nd and 3rd generation Kindle Scribes) which, in turn, makes colors look dull.

But increasing frontlight brightness to enjoy those colors means you’re draining battery life quicker. In the two months I’ve been using the Kindle Scribe Colorsoft, I’ve had to charge it every 2 to 3 days with an average use of 5 hours a day — mostly reading, but some writing too, and browsing the Kindle Store — with auto-brightness settings and Wi-Fi on. In comparison, the 2025 Kindle Scribe lasted me about 5 days on a single charge with the same settings and similar use case.

Annoyingly, the Colorsoft edition takes much longer than the standard Scribe to recharge as well.

Color isn’t always necessary

If color reading really is important to you, you’re better off with an iPad or a Samsung tablet to be honest — color E Ink just isn’t there yet. Moreover, most of us who use an ereader don’t necessarily even need color; we tend to mostly read text-heavy books and note-taking also, similarly, is mostly black ink.

On the Kindle Colorsoft Scribe, the color ink isn’t even as dark as I’ve seen on the Boox Note Air 5C, so it really doesn’t offer as much value for your hard-earned cash compared to some of its competition.

In comparison, its monochrome sibling is the overall better performer — as I’ve said in my Kindle Scribe (2025) review, it’s arguably the best digital notebook Amazon has made to date and, if it wasn’t so expensive, it would have scored full marks.

If you want a note-taking Kindle, I would still recommend the 2024 release of the Kindle Scribe — you might as well pick it up as long as it's still available. I suspect Amazon will discontinue it soon, but it’s a heck of a lot cheaper even at full price, is often discounted even outside of major sales and, according to Good Ereader, it will get the new software features that released with the current Scribe models, including Google Drive and Microsoft OneDrive connectivity, AI search and the refreshed homescreen layout.

If you’re willing to explore other E Ink writing tablets, I’d suggest looking into the reMarkable Paper Pure or even the Boox Go 10.3 series if you prefer an Android device that will also double as an ereader. A word of caution on the Boox devices, though: you'll need to deal with some ghosting.

One of the most exciting areas in technology is quantum computing, but it's also one of the areas that's been among the least accessible, difficult to relate to the real world, and still largely buried away in the lab. To reach the point at which quantum computers become useful, there's still plenty of work to be done.

The race to a million qubits

There are, and continue to be, different approaches to building quantum computers as well as all the components that power them. But the goal among all companies and institutions is the same: to build a machine that can outpace the fastest classical computer.

The cofounder and CEO of quantum computing company PsiQuantum, Jeremy O'Brien, was sharing his vision behind his organization, essentially saying that his team had attempted to work against a forged consensus.

Rather than scaling up qubits from a handful to 100 or 1,000, he said at the time that he was "fanatical" about getting to a million-plus qubits – the point at which experts expect there to be a meaningful performance leap.

New architectures and approaches

Intriguingly, O'Brien's comments came at a time when sheer scale and qubit count were the most sought-after elements in a quantum computer.

Since that point, so much research in the space has shifted not to the sheer size of the chips and how many qubits they can hold, but to building better qubits. Because quantum processors are so prone to failure, researchers have instead been trying to build processors with a few error-resistant qubits over millions of basic error-prone qubits.

The goal is still to achieve a quantum computer that can outpace classical machines, but experts expect that due to new breakthroughs, these quantum computers may need far fewer qubits than the millions we thought just a few years ago.

Last week the massive lawsuit against Meta, alleging that its products were designed to addict teens and may have caused them mental harm, went to court. Backed by dozens of US States, the potential $1.4 trillion price tag — if Meta lost — would have been enough to bankrupt the company.

That case is now, suddenly, over. Settled for the princely sum of $18 billion and promised fundamental changes in how Meta's major products, namely Instagram and Facebook, operate when used by anyone under 18. I'll get to the proposed changes below (which Meta would like to see YouTube and TikTok adopt, too), but first let's talk about the stunning turn of events.

We were in week two of the big trial, and honestly, we were just getting to the good part. After the states initially presented their case and made it clear they believed Meta knew its products might be harming teens and wasn't doing anything about it, Instagram CEO Adam Mosseri took the stand and was peppered with questions about the platform's current controls, like its prompts imploring teens to "take a break." The focus was so intense that the normally polished Mosseri appeared "annoyed." He also admitted that there are "no silver bullets" for safety.

On Wednesday (August 26), we were, it appeared, only hours away from the big show: Meta Founder and CEO Mark Zuckerberg taking the stand. I expected the prosecution to press him hard to see if they could trigger a slip-up or some reaction from the normally unflappable Zuckerberg.

Now, that moment will never come. Without admitting any wrongdoing or fault, Meta has closed the potentially business-killing or at least hobbling case.

Here's a look, though, at what Meta has agreed to in the settlement[PDF]. On the surface, it will sound like a lot. In fact, the changes are so sweeping that some on social media wondered if they could or should be applied to everyone on the platform.

• Time limits (2 hours)
• Mandatory pauses after 15 minutes of scrolling
• Overnight usage blocks
• No nighttime notifications
• No use during school hours
• No visible likes
• No cosmetic filters
• No algorithmically organized feed (if they choose)
• Age verification
• Better harmful content reporting
• Better parental controls

There are caveats: the two-hour time limit applies across Instagram and Facebook, though how much time does a kid under 18 spend on Facebook?

Likes removal includes any information that could lead to social comparison — essentially a FOMO filter.

Some of the changes, like age verification, parental controls, and content safeguards, represent strengthening and broadening of existing controls.

Meta has rolled out a whole suite of parental controls, especially on Instagram, in the last two years, but as Mosseri noted, not everything is working.

When I posted the above list on social media, one savvy reader made a salient point.

On Bluesky, Paul Dempsey posted:

Sadly, that needs at least these assumptions to hold:

1. Meta will apply tough-enough processes to stop kids hacking these restrictions.

2. Parents will be able to tell when their children have hacked the barriers.

3. States will have enough resources to police the new format.

I'm thinking 0/3.

A steep climb and long odds

He's right, of course. Meta's tools for blocking wily teens have not necessarily borne fruit. If anything, teens are turning away from social media of their own accord, but that doesn't mean millions aren't still on it and are often subjected to unhealthy content and social feedback.

Meta's decision to use AI to detect teen-like behavior and then automatically shift these accounts into the current teen account restrictions has worked for some, but how well does the AI really understand today's teens, and all the ways they can model adult behavior to get around the restrictions?

With this settlement, Meta will be cooking up stricter rules that may include some form of ID not just for the addition of teen accounts but some proof that people are who they say they are and that they're the parents of these newly onboarded teens.

Platform anonymity could take a hit or die altogether.

But as Dempsey noted, it's unclear if Meta can come up with enough ways to block resourceful teens.

His second question is also valid. So many parents are out of the loop when it comes to their children's and teens' online activities. Parents may be on Facebook but are less likely to casually swim in Reels and TikTok waters. Teens' comfort level with these systems far outstrips their parents'. I often wonder who will be leading whom as young people and their parents try to navigate these new restrictions.

Dempsey's last point is probably the ringer. I applaud all the states for joining together to pursue this one case. It's an amazing level of cooperation. Still, when it comes to laws and penalties, I'd be shocked if we saw even a whiff of consistency. That means the rules for policing, detecting, enforcing, and prosecuting those who skirt them will differ in every state. Worse yet, none will have the infrastructure or resources to keep track.

Sure, the rules across Instagram, Facebook, and other social media platforms that join will be consistent, but the tactics for circumventing them will be varied and no one — not the parents nor the states that originally brought the case — will know, that is until a teen reports harm or parents notice that their children are somehow engaging on social media with adults (or worse).

For Meta, these changes will alter their products, but I don't think that'll materially harm it or its business. It's a lot of money, but a manageable sum (the cost of one acquisition). For parents and teens, they will all struggle with the platform changes until their teens figure out workarounds, and parents again wonder why it's gotten so quiet in their kids' bedrooms.

A lot has already been said about the doom and gloom surrounding AI - some of it drawing on the warnings of classic dystopian novels, but much of it grounded in reality. There is a lot of optimism and promises from those in control of the technology, but not much has emerged in the form of actual benefit.

“I think by far the most accurate criticism of AI companies including Anthropic is that we haven’t yet delivered on our big promises to benefit the world,” Anthropic CEO and co-founder Dario Amodei has said in a post on X.

The technology is suffering from a “crisis of trust” that stems from people’s belief that companies and governments are “cooking up some new way to screw them over,” he argues. But given how much of the US military, law enforcement, and intelligence gathering infrastructure now relies on AI, are the people wrong?

AI can’t public relations its way out of this one

As AI technology took off, much of the benefit of the technology was aimed at convincing executives and C-suites that the technology - while unproven - was a worthwhile investment. As a result, multiple AI leaders positioned the technology as a way to strip away the labor costs of companies by replacing workers with AI.

Some AI gurus are still arguing these points behind closed doors. I attended a private event this year where an AI "guru" predicted that AI superintelligence would erase as much as 50% of the global workforce within 10 years, and the companies best positioned to take advantage of this cataclysmic event would reap unimaginable control and riches. This was naturally met by applause from the upper echelons of management from companies in attendance.

Amodei himself famously said that AI could wipe out 50% of white collar jobs. But he has since restructured this message, saying instead that AI is a ‘multiplier of output’ - or corporate word salad for ‘increasing productivity without having to increase compensation’.

As I have said numerous times, the proof of AI really is in the pudding. We’ve heard everything that it could do, but seen little benefit in what it can do. To his point, Amodei acknowledges that.

He notes that if an advert went up saying that AI could cure cancer, it would be perceived as entirely misleading. “The thing that will work is actually curing cancer,” he says.

Amodei has teased Anthropic’s morality on these issues, saying that the company is “ramping up its efforts” to focus on using AI for biology and medicine. But again, we’re yet to see the results.

Amodei’s post also touched on where AI is centralizing power, echoing some of Mark Zuckerberg’s recent rant on how best to use AI for everyone’s benefit. Amodei’s argument differed slightly from Zuckerberg’s, arguing that there is a balance to be struck between regulation, oversight, testing, and open-weight models.

He notes that in some aspects he agrees with the “Pacing the Frontier” movement, stating that implementing controls on the frontier labs would “modulate” the pace of the technology while simultaneously allowing the rest of the industry to catch up.

The obvious flaw in this plan is that no other nation would be subject to this regulation which could deliberately hamstring US hegemony in the AI industry - something the Trump administration has frequently said it will not allow to happen.

Is AI delivering more benefits to authoritarian infrastructure than to people?

Where AI systems have been deployed extensively is in surveillance infrastructure. Until the collapse of its relationship with the Pentagon, Anthropic was working with the Department of Defense to procure AI systems. Anthropic refused to cooperate in allowing its technology to be used for mass surveillance and fully autonomous weapons, and lost the contract.

Luckily for the DoD, OpenAI shook hands on a new deal with the Pentagon and immediately filled the shoes left by Anthropic with a prohibition on its technology being used to spy on Americans.

That doesn’t mean other companies are abiding by the same level of ethics. Flock has faced a widespread guerrilla movement targeting its AI-powered cameras designed to track vehicle number plates across the US.

The system is intended to flag suspect’s cars as they travel across boundaries and lines of authority, but has instead seen success as a way for members of law enforcement to stalk exes, monitor colleagues, and track their love interests.

AI systems are also being deployed in traffic monitoring and speeding cameras. This in particular has received a lot of attention as a grey area of policing, especially for systems that misidentify numberplates or are wrongly configured.

For criminal cases in the US, the burden of proof rests with the government to prove someone committed a crime. However, some autonomous ticketing systems that have issued citations to the wrong vehicles have been criticized for requiring those accused to prove their innocence in court, flipping the burden of proof for criminal cases on its head.

And then there is the elephant in the room.

Palantir’s AI technology is deployed across the US military and law enforcement. Most notably, it has been used by the US Immigration and Customs Enforcement (ICE) to track those suspected of illegal immigration - but Palantir’s technology is also used in drone imagery analysis, long-range precision strikes, and military data analysis and planning.

No doubt it has been instrumental in delivering the World Record for the most times a single war has been won.

AI is great news if you have nothing to hide and love being on the side of the Good Guys™, but that's a constantly shifting definition.

Imagine AI is the sun and businesses are Icarus. By now, most of us are basking in the light of this transformative innovation, which sometimes feels too hot and other times disappears behind clouds so dense we wonder why we ever trusted it in the first place.

Some intrepid businesses not only gaze at AI's warm light, but openly worship it. They don wings, and fly as close as they can to its surface. The goal: scoop up a bit of that generative magic, and fly back to Earth with agents, which, to carry this insane metaphor further, can be fashioned into people-like workers capable of doing what we do, but at scale and with little-to-no monitoring.

Of course, the flight to AI's core is dangerous, and so we have episodes like Meta's recent Quixotic adventure which, according to an extensive Reuters report, saw it attempt to transform a vast portion of its workforce (reportedly 60%) into AI agents overseen by a much smaller set of highly skilled workers.

That Meta's wings almost immediately started smoking from its proximity to the AI core is no surprise, at least not to me.

Slow down

I talk and write about AI a lot, and I've been covering it in some form or another for almost 25 years, back when companies like IBM and Microsoft were the AI leaders, and we thought Deep Blue beating Kasparov was the apex of what we could expect from artificial intelligence.

Modern generative AI burns far brighter, and shows in real time such vast potential that enthusiasm is not just rampant; it's out of control.

AI agents, which Meta apparently started using at scale, are driving much of this mania. These automated tools offer the promise of handling drudgery-filled tasks in innumerable businesses, and could solve not just productivity problems, but be the answer to workforce issues, including hiring problems, underperformers, and burnout. Agentic AI can not just handle repetitive tasks; it can cobble together disparate systems to complete work.

From Meta's perspective, especially as a company that is primarily a software business, AI that could work across tools and systems to not only help keep things running but build new work (under the supervision of a smaller set of super-performers) probably seemed like the perfect solution.

Conclusions like this are not exclusive to Meta or companies of its size. Conversations about the use of AI in business persist in every industry and at every level.

AI has gone to work in healthcare, retail and e-commerce, manufacturing, financial services and banking, entertainment, and media. All with varying degrees of success and implications.

It's a stunning fact when you consider that we only met OpenAI's ChatGPT in 2022. Never has a technology so quickly infiltrated every facet of life, and so consumed daily conversation. AI Time is triple time (3x the speed of any previous innovation), and that includes adoption.

There is, at least in business, a shocking level of trust in what is still, in many ways, an untested solution. I'm not saying AI does not work, but there is the assumption right now that it works in every situation.

Certainly, if the Reuters report is accurate, and Meta essentially verified 90% of it (it claims it was never planning on laying off 60% of its workforce), Meta is guilty of this.

Do you smell something burning?

With AI, the act of implementation and business transformation tend to happen in parallel, so as Meta was planning and enacting its reorganization and layoffs, it was already employing Agentic AI.

"Infrastructure teams were also warning of 'reliability warning signs' caused by the AI coding surge, according to an internal post," wrote Katie Paul for Reuters.

There was, the author adds, evidence that these issues (and basically faulty work) were visible in June when hackers used Meta's AI chatbot to hack celebrity Instagram accounts.

Between that and what appears to be a company revolt (my favorite part was the bathroom stall notices encouraging people to protest Meta's use of keystroke tracking), it was clear Meta has gone too far, too fast with AI for everyone's tastes.

In the end, with its feathers on fire, Meta retreated from the AI sun (or at least its plan) and canceled layoffs and even undid some of the reorg.

The lesson here is not that AI is too dangerous to approach closely, or that AI and business are a bad match. But, as we are seeing, pronouncements that AI would transform businesses without removing or replacing jobs were misinformed, and allowing AI to produce work without adequate human oversight at every stage of the process is, for the time being, ill-advised.

As Mark Zuckerberg tries to rewrite his own script to be more people-first, it might be time for everyone to take a beat, and reassess how AI fits into their businesses, workflow, and life.

AI might feel like the sun, but it can't be our sole source of light and warmth. It's still just another tool — sure, a powerful and incredibly fungible one, but still a tool — and over-reliance on it will undoubtedly lead to frustration, or worse.

Meta learned a hard lesson: your AI plans must be grounded in a human-centric approach; otherwise, the humans will rise up and clip your wings.

Scary AI spend and weak access security controls are the flipside of the sharp rise in agentic AI. The answer lies in the form of a control plane that provides the visibility CIOs need.

The past couple of years have seen enterprises press the fast-forward button on generative AI strategies. e have all seen a remarkable change in deployment patterns from useful chatbots to autonomous agents that engage deeply with core software, data and LLMs, and even create their own offspring agents.

CIOs and others are rightly excited about the capacity of agentic AI to reduce manual processes and accelerate speed to insight and actions.

However, governance and AI token costs are alarming challenges.

What’s needed today then is the power of AI without the major concerns over incurring access control risks relating to ‘Shadow AI’ and agent sprawl, and without the ‘bill shock’ relating to unmanaged token consumption.

Moving fast

The generative AI shift that has occurred so quickly spans the early days of ad hoc AI experimentation via testing and pilots to today’s enterprise-grade business processes. It has also outpaced the progress of the usual governance frameworks and effectively rendered them obsolete.

Experienced CIOs may compare the change to the early days of cloud computing platforms when staff with credit cards expensed the AWS trials they had spun up and quickly ran up staggering bills.

Or they may hark back to the days when sales leaders set up online CRM platforms or other process owners bypassed IT to run their own Sharepoint systems.

Or, to reach for yet another lesson from history, they may compare what is happening with generative AI now with the advent of bring-your-own-device schemes and WiFi.

Remember when the user-led demand to have the shiny new gadget ran a coach and horses through the command-and-control IT regimes of cloned PCs stamped with a single image and CIOs no longer managed everything that touched their systems?

The problem is here. Deal with it

However we make the comparison, it’s clear that today there are very real risks relating to data leakage, compliance standards and IT costs because of the speed with which AI has advanced. Nobody wants to lose a chance to gain a competitive advantage and even to fundamentally re-engineer and automate expensive, error-strewn manual processes. So CIOs are once again being asked to square a circle.

As IDC has said: “There is a well-established playbook for calculating the return on an IT investment. Define a baseline, model the efficiency gains, project the cost savings, and present a number to the CFO. For most technology investments, that approach works. Agentic AI breaks it.”

Agentic AI is not fixed and predictable so it’s tough to manage and calculate return on investment, but we know it can take out swathes of cost, so we have to use it and we have to manage costs as best we can.

Gartner has already suggested that by 2027, more than 40% of agentic AI projects will be cancelled due to factors such as cost overrun fears or inadequate risk management. That’s not good enough. Corporate IT must step up and deliver the value the business needs, just without the attendant risks and costs.

The solution here is to head off undesirable outcomes at the pass by providing a check, a little like an airport security system that identifies passengers and scans for any potentially unsafe activity or nearby environmental factors such as possession of sharp objects.

You will see this happen in IT very quickly now, such are the risks and speed of what is happening. If 2026 is the year of agentic AI, then it also has to be the year of secure agentic AI.

Enter the AI gateway

One way to address the challenges outlined above is to create a gateway or filter. This approach offers a single control layer regardless of whether the AI model is hosted in public or deployed as a private, self-hosted entity. It effectively secures traffic from agents to LLMs and from MCP servers. This in turn enables organizations to have centralized policies and a clear view of token accounting.

It even has the visibility impact of informing decisions of non-techies who know little about AI operations. So, for example, a CIO working in tandem with a CFO perhaps may identify a better cost efficiency (for example, switching from a hosted model to a private model) and act accordingly.

That’s great news for CIOs who want to maintain a reputation for IT being agents of change but not its less desirable reputation for being risky and pricey.

Looking for the best cloud storage? These are our top picks.

This article was produced as part of TechRadar Pro Perspectives, our channel to feature the best and brightest minds in the technology industry today.

The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/pro/perspectives-how-to-submit

Autonomous AI agents are moving faster than the frameworks meant to contain them. Enterprises are deploying agents that can call systems, pull data, and increasingly interact with other agents to complete multi-step tasks.

Yet few organizations can say with confidence exactly how many agents are running in their environment, what each one is authorized to touch, or who exactly is accountable when something goes wrong.

That gap can turn agentic workflows from a promising technological advancement into a potential minefield of risk unless enterprises take a new approach to governance – one that provides greater oversight into how agents are operating and what systems they can access, trust, and use.

Don’t wait for perfection

The market has responded to the potential threat around AI agents going rogue with a wave of new tooling: agent discovery platforms that scan for active agents, and agent harnesses that box them into approved boundaries. Both are useful, but neither solves the problem alone.

The challenge of “getting a handle” on AI agents is compounded by the pace of change. New agent tools, new AI model releases, and new orchestration options are arriving at a rate that makes any static governance model obsolete within months.

Rather than waiting for a “perfect” governance framework to emerge as a standard, organizations should take steps now to create a working structure that can evolve over time.

So, what might this look like in practice?

Continuous visibility and granular guardrails

Policies and procedures alone cannot confirm what is actually running in production. Organizations need a registration and discovery process that captures every agent in use, not just the ones teams report having built.

Real visibility comes from instrumenting the environment itself, using logging and observability data generated by the underlying models, and building analysis on top of it. Only that raw data can show what an agent is actually doing, how often, and at what cost, rather than relying on what people think it is doing.

Another key step is to move from broad guardrails to granular ones. A single, broad-based harness applied uniformly across an organization, for example, limits scope without addressing risk in a meaningful way. Different use cases call for different levels of restriction, and treating every agent identically either over-constrains valuable work or under-constrains risky work.

Guardrails need to be defined at the level of the individual agent and its specific task, not the organization as a whole. As agents begin accessing other agents to complete a task, that specificity becomes even more important: each agent needs an explicit, narrow definition of what it can do and what it can reach, so that any leakage beyond that defined scope is immediately visible.

With all of the above, enterprises should embrace an approach of “validate, don’t assume”. Putting a conceptual harness or policy in place is only the first step. Confirming that it actually holds under real conditions is a separate and ongoing exercise.

A useful approach here is to pair conceptual guardrails with a logical or physical enforcement layer: a tool that can confirm what is executable, what an agent can interact with, and whether it stays within that boundary in practice.

Continuous monitoring against policy, not a one-time sign-off, is what proves a harness is working, because agent capability and the surrounding set of tools change on a near-daily basis.

Access is not the only variable

Governance conversations tend to focus almost exclusively on what an agent can access. Equally important is the order in which it accesses information. An agent that gathers information out of sequence, or acts on data before a dependent step has completed, can produce results that are wrong even if every individual action was technically permitted.

Some workflows require steps to run in parallel; others require specific sequencing to ensure the most accurate results. Building a discrete, well-defined operational sequence for each use case, rather than allowing a model to determine its own order of operations, substantially reduces this type of error.

This is as much a process design issue as a technical one, and it relies on people defining the use case clearly before any tooling is applied.

A disciplined approach reins in agentic risk

None of these elements works in isolation. Discovery without granular guardrails leaves organizations knowing what exists but not controlling it. Harnesses without validation create a false sense of security. And governance focused only on access misses key variables that can introduce risk.

The organizations managing this well are the ones treating agent governance as an extension of existing IT and data governance, applying the same discipline used for foundational legacy systems rather than treating it as a bolt-on afterthought.

Given how quickly the technological landscape continues to shift, this comprehensive governed approach is what will keep agentic AI delivering value – and reduce the risk of agents going rogue.

We've featured the best AI tool.

This article was produced as part of TechRadar Pro Perspectives, our channel to feature the best and brightest minds in the technology industry today.

The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/pro/perspectives-how-to-submit