On May 17, 2026 Mistral AI co‑founder and CEO Arthur Mensch told a French commission of inquiry that contemporary large language models “can orchestrate attacks, detect software vulnerabilities and propose concrete exploits.” He said those technical capabilities turn routine code analysis into a novel cyber risk that requires urgent attention from policymakers and defense planners.
Mensch cited Anthropic’s Mythos as an example of a model able to analyze code and surface exploitable flaws. He warned that allowing external AI systems to process sensitive military software would risk creating a durable dependency that would be difficult to reverse, and stressed that comparable vulnerabilities could be exposed by other vendors’ models, including Mistral’s or models developed in China.
His testimony came as EU officials negotiate early access to advanced cybersecurity models from OpenAI and Anthropic. Mensch framed the issue as a policy trade‑off: granting privileged access to foreign AI tools could deepen technical dependence and shape procurement, incident response and national‑security posture. That forces France and the EU to weigh short‑term defensive gains from established providers against the strategic goal of nurturing European alternatives.
Mensch also addressed Mistral’s finances, saying U.S. investors own under 30% of the company and that European capital — while preferred — is currently insufficient. He said a sale is not planned; instead, Mistral intends to remain independent, pursue an eventual initial public offering, and position itself as the only EU firm with competitive large‑language models.
Sources
Replies (0)
No replies in this topic yet.
