Ex-MI5 Officer's AI Security Startup Overmind Raises $2.5M Seed

Jan de Vries · 2026-02-11

Here's something that caught my eye recently. Overmind, a London-based startup, just announced a $2.5 million Seed funding round. They're building what they call the "supervision layer" for AI agents. Think of it like a lifeguard for autonomous AI systems—watching, ready to jump in if things start to go sideways. They plan to use this capital to grow their technical teams and accelerate product development. Their initial focus? Highly regulated sectors like legal, healthcare, and FinTech. That's where AI agents promise huge efficiency gains, but the stakes for mistakes—and the rules around data privacy—are incredibly high. ### The Investment Backers The round was led by Osney Capital, a specialist cybersecurity investor. They weren't alone. A group of other venture firms joined in, including 14Peaks, Portfolio Ventures, Antler, and Endurance Ventures. That's a pretty solid vote of confidence from people who know the security landscape inside and out.  ### A Different Take on AI Security Here's where it gets interesting. The CEO, Tyler Edwards, has a provocative take. He argues the industry is focused on securing the wrong thing. > "Models will always be vulnerable to adversarial inputs – that's a fundamental property of how they work," Edwards says. "But what happens when an agent is live in production, interacting with real systems, and its behavior starts to drift? Right now, most teams have no idea." His point is simple but powerful. You can have the most secure model in the world, but once it's out there making decisions on its own, new risks emerge. Overmind's tech is built to monitor those live interactions and intervene before any real damage is done. It's security for the deployment phase, not just the training phase. ### The Broader European Funding Wave Overmind's raise isn't happening in a vacuum. It's part of a significant wave of investment flowing into European startups tackling AI agent security and governance. Just look at the activity in 2025-2026: - **Archestra** (London): Raised $3 million in pre-Seed funding to build safety guardrails for connecting AI agents to internal data. - **Equixly** (Italy): Secured a hefty $10.8 million to scale its AI-driven API security testing platform. - **Qevlar AI** (France): Landed $9.8 million to develop agentic AI specifically for security operations centers. And that's just the direct players. Other companies in adjacent spaces, like those focusing on AI analytics for brand safety or agent-driven marketing platforms, have also pulled in millions. All told, we're looking at over $28 million recently invested in this niche. The message from investors is clear: as AI gets more autonomous, the tools to supervise it are becoming a critical, and valuable, market. ### The Team Behind the Tech What really sets Overmind apart might be its founding team. They've blended deep, real-world security experience with a track record of scaling tech companies. CEO Tyler Edwards spent eight years building AI systems for British intelligence agencies—MI5, MI6, and GCHQ. That's about as "in the trenches" as you can get for high-stakes security. His CTO, Akhat Rakishev, previously led machine learning infrastructure at major companies like Monzo and Lyst. Meanwhile, the CRO, Sam Brunt, has scaled go-to-market strategies at not one, but three unicorn startups. It's a combination that gives them unique credibility. They're not just theorizing about security threats; they've lived them. And they know how to build and sell enterprise software. ### Why This Matters Now Adam Cragg from Osney Capital put it well. He called agent security, performance, and execution "the ultimate competitive advantages" in the new frontier of autonomous AI. Overmind's tool doesn't just watch for problems; it uses what it learns from production data to iteratively improve the AI's performance. It's a feedback loop that aims to make agents both safer and smarter over time. Another investor, Adam French from Antler, believes the team is "defining the security standard for how superintelligence will be safely deployed in production." That's a bold vision, but it shows the level of ambition here. So, what's the takeaway? We're moving beyond just building powerful AI. The next big challenge, and opportunity, is building the trust and control systems that let us deploy it safely at scale. Startups like Overmind, backed by experienced teams and savvy investors, are betting they can provide that essential layer of confidence.