Mistral CEO Warns France Against Using Anthropic’s AI for Military Code Scans

Mistral AI CEO Arthur Mensch has urged France to avoid using Anthropic’s Mythos artificial intelligence models to analyze military code repositories, warning of cybersecurity vulnerabilities and growing reliance on U.S. technology, according to a July 2025 report from The Decoder.

Mensch emphasized that modern AI systems—including those developed by Mistral—could potentially identify software exploits or orchestrate cyberattacks, making their use on sensitive infrastructure “inadvisable.” The caution comes as France evaluates AI tools for defense applications, with Mensch ruling out Mistral’s participation in such contracts amid the company’s focus on an upcoming initial public offering.

The warning highlights tensions over the deployment of U.S.-based AI models in European defense systems. Anthropic, the San Francisco-based developer of Mythos, has not publicly commented on the restrictions. The issue underscores broader debates about data sovereignty and the risks of foreign tech in critical sectors.

France’s military code-scanning initiative had previously considered multiple AI providers, but Mensch’s stance reinforces calls for European nations to prioritize locally developed models for sensitive work.