Escape Raises $18M to Build AI Agents for Live Security

Escape, an offensive security engineering platform, has secured $18 million in a Series A round to develop AI agents aimed at automating the security lifecycle in live environments. The round was led by Balderton, with participation from Uncorrelated Ventures and existing backers IRIS and Y Combinator.

Founders Tristan Kalos (CEO) and Antoine Carossio (CTO) started Escape to tackle gaps in traditional application security — particularly the risks that only appear in production when integrations, configurations and authentication flows are live. The startup says legacy scanners and manual penetration testing often miss logic flaws and data exposure that manifest after deployment.

Escape’s approach, which it calls offensive security engineering, uses AI agents that simulate attacker behaviour to discover attack surfaces, run continuous security tests and suggest or implement remediation. These agents are designed to work directly within engineering workflows so teams can move more quickly from detection to resolution while lowering operational overhead.

In a recent analysis, the company reported finding over 2,000 high-impact vulnerabilities across 5,600 publicly available apps generated by automated coding tools — including 175 instances of personal data exposure and multiple cases of leaked credentials. Escape says these issues existed in live production systems and were discoverable within a short timeframe.

The fresh capital will go toward improving the platform’s agent capabilities — including tools to analyse application logic during penetration tests — and scaling engineering and go-to-market teams as the startup targets enterprise customers across the US and Europe. Kalos said teams are struggling with siloed, manual processes and need automated, scalable solutions as both coding and attacks accelerate.

For security teams facing increasing velocity and complexity, Escape’s model aims to shift more of the detection and remediation burden into automated agents that operate where risks actually appear: in production.