RevEng.AI

Earlier this month, a small British team released two of the most capable AI models of their kind in the world, and it deserves a little more attention. There is a lot of debate about whether Europe can build frontier AI. Amongst the noise, we have been quietly training specialised models from scratch in London, and they now outperform the best tools available anywhere in their field. These are not the first foundational models built in the UK, but they are some of the most ground-breaking. They are called Ventris and WilBERT, the first two releases in our suite of models for binary analysis. What is binary analysis? Almost all software arrives as a compiled binary - the ones and zeros a computer processor executes. The source code a developer wrote is effectively gone - you can run the program, but you cannot read it. That is fine until you need to know what a program is really doing. Is this file malicious? Does this supplier's software contain hidden flaws? Usually nobody knows, because reading compiled software is slow, highly skilled work. Ventris translates compiled software back into readable code. WilBERT recognises code it has seen before, even when rebuilt for a different processor, which is how you spot known malware, reused components and unpatched vulnerabilities hiding inside something unfamiliar. A capable reverse engineer analyst costs £1,500 to £2,000 a day, and a single file can take anywhere from a day to a month to inspect. With our models, analysis that previously cost thousands and took weeks now costs pennies and takes minutes, which means verifying a whole software estate rather than triaging a handful of files a quarter. How the models perform The benchmark results have been extraordinary. Ventris substantially outperforms the established commercial and open-source decompilers, and matches or outperforms frontier model families at up to 1,000 times lower cost per function. Frontier models from Anthropic and OpenAI are extraordinary engineering, but they are built to do everything, and at that scale analysing every function in every binary makes no economic sense. Sovereignty by default Our models are light enough to run on premise on modest hardware, so the software you analyse never leaves your estate. That matters for defence, government and regulated organisations who cannot send proprietary binaries to a third-party API. 🇬🇧 A great moment for British AI This is an important step on a long journey. State of the art models, trained from scratch by a small team in London, standing shoulder to shoulder with giants. And we're just getting started... Thank you to our investors, partners and customers, and above all to the incredible team who built this - we could not have done this without you. 🔗 Full technical write up in the comments. Stop trusting. Start verifying. #BritishAI #BinaryAnalysis #SovereignAI

Today, I’m proud to formally introduce Mega Bite, RevEng.AI’s suite of specialized AI models for binary analysis. This includes two new proprietary models, WilBERT and Ventris, built to analyze binary software without requiring access to source code, closing a visibility gap that traditional decompilers and general-purpose LLMs have never managed to. This launch represents an enormous amount of research, engineering, and persistence from the RevEng.AI team, and I’m incredibly proud of what we have built together. As organizations rely on more third-party, commercial, and AI-generated software, understanding what is actually inside each executable has never been more important. WilBERT and Ventris change what is possible in binary analysis, giving security teams the speed, accuracy, and confidence needed to verify software at scale. https://lnkd.in/ejsTNt53 #RevEngAI #BinaryAnalysis #SoftwareSupplyChainSecurity

Today, I’m proud to formally introduce Mega Bite, RevEng.AI’s suite of specialized AI models for binary analysis. This includes two new proprietary models, WilBERT and Ventris, built to analyze binary software without requiring access to source code, closing a visibility gap that traditional decompilers and general-purpose LLMs have never managed to. This launch represents an enormous amount of research, engineering, and persistence from the RevEng.AI team, and I’m incredibly proud of what we have built together. As organizations rely on more third-party, commercial, and AI-generated software, understanding what is actually inside each executable has never been more important. WilBERT and Ventris change what is possible in binary analysis, giving security teams the speed, accuracy, and confidence needed to verify software at scale. https://lnkd.in/ejsTNt53 #RevEngAI #BinaryAnalysis #SoftwareSupplyChainSecurity

Something big is coming from RevEng.AI — a breakthrough that's about to redefine what's possible in binary analysis, and the wait is almost over. Watch this space! 👀 🚀 #ReverseEngineering #CyberSecurity #AI

A huge thank you to HCLSoftware for having us yesterday and hosting an amazing event. Some extremely insightful conversations, great pitches, and a genuinely picturesque venue at Wormsley Estate (as you can see!). Shame that after 40 days of sun it decided to rain sporadically, but that didn't dampen what was a really successful event, and we look forward to continuing the conversations over the coming months. #CyberSecurity #Innovation #thirdpartyrisk #supplychainsecurity #softwareassurance #TechEvents #HCLSoftware

We're thrilled to be part of Innovation Exchange 2026 with HCLSoftware. We're looking forward to showing attendees how RevEng.AI helps businesses catch tampering, hidden functionality, and backdoors introduced through third-party components, builds, or shipped packages. The threats source-level review alone can't see. See you there! #InnovationExchange2026 #CyberSecurity #SupplyChainSecurity #BinaryAnalysis #HCLSoftware #Innovation #malwareanalysis

Reverse engineering has always forced a tradeoff: speed or depth. Not anymore. Every unidentified binary sitting in a queue is both a technical bottleneck and a business risk, the longer it takes to understand what a binary actually does, the longer your exposure window stays open, and the more your best analysts get stuck doing repetitive triage instead of real investigative work. RevEng.ai uses AI to close that gap. Automated function identification. Behavioural analysis at scale. Faster cuts through obfuscation and stripped binaries. Matching against known vulnerabilities and malware families in minutes instead of days. The result isn't just faster analysis, it's faster remediation, lower risk exposure, and a team that can finally focus its expertise where it counts instead of burning hours on repetitive groundwork. Whether you're the one measuring risk or the one in the binary, the outcome is the same: less time waiting, more time acting. Curious how this fits into your security stack? Let's talk. #ReverseEngineering #CyberSecurity #MalwareAnalysis #AI #InfoSec #RiskManagement #IncidentResponse

We're at Black Hat Vegas right now, come say hi! The RevEng team is on the floor this year, and we've got a limited stash of hats to give away to the lucky few we cross paths with. DM us or drop a comment and let's grab a minute to chat reverse engineering, swap war stories, or just meet the team in person. First come, first served on the hats — they won't last long. See you out there 🖤 #BlackHat #BlackHatUSA #BHUSA #RevEng #ReverseEngineering #InfoSec #CyberSecurity #Vegas

A skilled reverse engineer can spend hours (sometimes days) unpicking a single obfuscated binary. Multiply that across the volume hitting teams daily, and "we'll get to it" becomes the default response to a threat that's already moving. This is where AI earns its place — not as a replacement for the analyst, but as leverage. Semantic function matching, automated similarity detection against known malware families, and rapid triage of what's actually novel versus what's a repackaged variant, this is grunt work AI does in seconds that used to eat a working day. That frees the human for what humans are still best at: judgment calls, novel techniques, and the "wait, this doesn't add up" instinct no model has yet. At RevEng.AI, this is exactly the gap we're closing, using AI to compress the time between "we have a sample" and "we know what it does." Not hype, just hours back to your analysts' day. #CyberSecurity #InfoSec #MalwareAnalysis #ReverseEngineering