Agentic AI is moving fast, but its risks are evolving even faster. As organisations deploy autonomous systems capable of making decisions and acting independently, traditional cybersecurity tools are struggling to keep up. These agents are vulnerable not only to adversarial prompts or corrupted data but to unpredictable behaviour once they begin executing tasks in live environments. This emerging gap is already slowing innovation.
Gartner warns that 40% of agentic AI projects may be cancelled by 2027 due to inadequate safeguards. In other words, without a fresh approach to security, the promise of agentic AI could stall before it reaches scale.
Intelligence-grade layer of protection
London-based Overmind is stepping directly into this void. The company has developed a supervision layer built specifically for the unique challenges of agentic AI, offering real-time oversight that goes far beyond standard monitoring tools. Its platform continuously tracks behaviour, spotting deviations the moment they occur and stopping harmful actions before a human operator can intervene. Beyond protection, Overmind strengthens agents over time. Using reinforcement learning, the system improves performance and accuracy, enabling agents to evolve into safer, more reliable tools for complex workflows.
To accelerate its mission, Overmind has raised a £2M seed round. The investment was led by specialist cybersecurity firm Osney Capital, joined by 14Peaks, Portfolio Ventures, Antler, and Endurance Ventures.
The funding will expand technical teams, intensify product development, and strengthen go-to-market operations. Early focus areas include legal, healthcare, and fintech sectors, where agentic AI could unlock massive efficiency gains but must operate under some of the strictest rules on data privacy, compliance, and risk management. By offering assurance that these systems can be deployed safely, Overmind aims to remove the friction currently holding organisations back.
Built by experts who understand high-stakes security
Overmind’s credibility is reinforced by the experience of its founding team. CEO Tyler Edwards spent eight years creating AI systems for the UK’s intelligence agencies, including MI5, MI6, and GCHQ, institutions where failure is not an option. CTO Akhat Rakishev previously built machine-learning infrastructure at Monzo and Lyst, while CRO Sam Brunt helped scale three unicorns, such as Funding Circle, Pipe, and Vertice. Their combined background blends national security, high-growth engineering, and commercial scale, giving Overmind a rare mix of technical depth and operational pragmatism.
Looking ahead
Overmind’s launch marks an important shift for the industry. Agentic AI may be powerful, but without embedded security, it cannot earn the trust required for widespread adoption. By bringing intelligence-grade protection to commercial AI tools, this new startup is positioning itself at the heart of the next major leap in enterprise technology, where autonomy and safety must develop hand in hand.
Tyler Edwards, Co-Founder and CEO of Overmind: “The AI security industry is trying to secure the wrong thing. Models will always be vulnerable to adversarial inputs – that’s a fundamental property of how they work. But what happens when an agent is live in production, interacting with real systems, and its behaviour starts to drift? Right now, most teams have no idea. Overmind provides the deployment-layer infrastructure needed to monitor agent interactions and intervene before damage occurs.”
Adam Cragg, Partner at Osney Capital: “In the new frontier of autonomous AI, agent security, performance, and execution are the ultimate competitive advantages. Overmind provides businesses with truly differentiated technology that monitors and secures agentic AI while iteratively improving model performance, enabling teams to scale with confidence. We’re excited to back such a strong founding team addressing a critical market.”
Adam French, Partner at Antler: “Overmind is addressing one of the most critical bottlenecks in the advancement of AI: the security and supervision of autonomous agents. The founding team is uniquely positioned to solve this and deliver the ‘intelligence-grade’ security AI tools needed. We are proud to back a team that isn’t just building a tool, but is defining the security standard for how superintelligence will be safely deployed in production.”
