Mistral AI Updates: May 18, 2026

1. Mensch tells a French commission to keep US frontier models out of military code bases

Mistral. Mistral CEO Arthur Mensch testified before a French commission of inquiry that Europe is now structurally dependent on US frontier models, and that letting Anthropic’s Claude Mythos scan French military code bases would compound that exposure. Mensch argued that modern frontier models can already orchestrate offensive operations end to end - finding vulnerabilities, planning attacks, and suggesting exploits - which makes the choice of which lab gets to read sensitive code a sovereignty question rather than a procurement question. He paired the warning with a renewed push for European AI capacity and pointed toward eventual public-market financing for Mistral, though he also stepped back from short-term IPO timing in line with his earlier denials of imminent listing plans. Source