Fleuret AI raises €4M to build agentic pentesting platform

Fleuret AI sits at an important intersection: agentic AI and security. Instead of using LLM agents just for customer support or coding, Fleuret’s platform uses them to simulate attackers, chain together exploits and validate impact, effectively turning the red‑team function into a continuous AI service.([fleuret.ai](https://fleuret.ai/news/fleuret-raises-4m?utm_source=openai)) As more infrastructure and business logic is wired through autonomous agents, the attack surface they create is growing faster than human security teams can keep up.

This kind of tooling matters for the AGI race because it is one of the first attempts to put “offensive agents” to work in a structured, controlled way on behalf of defenders. If it works, it could become part of the baseline security posture for any serious AI deployment, especially under regimes like NIS2 that require ongoing assurance rather than point‑in‑time audits.([fleuret.ai](https://fleuret.ai/news/fleuret-raises-4m?utm_source=openai))

It also signals that European capital markets are willing to fund very specialized AI startups that may never train their own frontier models but sit on top of them to solve high-stakes problems. That diversification of the AI stack is healthy: not every important AI company needs to be a foundation‑model lab to be strategically relevant.