France names Mistral sovereign AI provider for government cybersecurity, excludes OpenAI

France threw its weight behind sovereign AI, with Budget Minister David Amiel announcing the government will engage domestic providers such as Mistral — and explicitly not OpenAI — to test public services for security vulnerabilities. The decision follows a DGFiP (tax agency) breach that exposed roughly 700,000 taxpayers' data, making national control over sensitive security tooling a political priority.

The exclusion of OpenAI is the pointed part: it elevates data sovereignty and European control over raw capability as the deciding procurement criterion, a stance Mistral has long championed by pitching itself as the guardian of Europe's AI independence. Mistral also shipped product this cycle, releasing Agentic Search — a retrieval layer that reduces token use and latency — plus Shieldstral 1.0 and Mistral Medium 3.1 updates.

Competitively, this is a template other European governments may follow, and it directly counters the U.S. labs' government-market gains (xAI, OpenAI, and Google on the Pentagon's GenAI.mil). It also reflects the week's cybersecurity-and-AI theme from a policy angle: nations increasingly view AI security tooling as critical infrastructure requiring domestic control. This is an August 18 story, but the sovereignty theme remains highly relevant amid the week's security incidents. Readers should watch which other EU states adopt similar exclusions, the scope of Mistral's government contracts, and whether sovereignty requirements fragment the AI market along national lines.