Geordie AI bets runtime agent security becomes infrastructure; $30M Balderton Series A

Geordie AI is a London cybersecurity startup founded in early 2025 by Henry Comfort, formerly COO of Darktrace's Americas business through its $5B exit; Hanah-Marie Darley, formerly Darktrace's director of security and AI strategy; and Benji Weber, formerly a senior engineering director at Snyk. Its product gives enterprise security teams real-time visibility into which AI agents exist, what data and tools they can access, how they behave, and what risks they create.

The bet is architectural: traditional security tools were built for predictable, rule-based software, while AI agents make non-deterministic decisions and evolve over time. Geordie profiles agents, establishes behavioral baselines and runs anomaly detection, plus a runtime remediation suite called Beam that constrains agent behavior without blocking deployment - covering cloud, code and endpoints rather than just inputs.

The gap is real and measurable: 82% of enterprises already run AI agents but only 44% have policies to secure them. Geordie reported 1,300% ARR growth in the first five months of 2026 and won the RSAC 2026 Innovation Sandbox in April. In May 2026 it raised a $30M Series A led by Balderton at roughly $180M post-money - reportedly the largest European cybersecurity Series A to date - bringing total funding to $36.5M.

Customers include AlphaSense, where Geordie covers tens of thousands of agents, and Owkin, which runs hundreds of AI agents over more than 50 petabytes of biomedical data; Owkin's internal risk modelling attributed $12M-13M of prevented exposure to one proof-of-concept deployment. The company employs 37 people across London and New York. It is scaling, with no exit.