Mistral challenges US AI firms over safety regulation claims

Mistral and a coalition of European AI companies are pushing back against calls from US firms to slow down frontier AI development, arguing the safety-first framing is little more than competitive strategy dressed up as public interest.

The French AI startup, now valued at roughly €21 billion, directly targeted Anthropic and other American AI leaders who have advocated for deceleration and coordinated safety measures.

The safety smokescreen argument

Anthropic CEO Dario Amodei published an essay on September 12 advocating for voluntary coordination among AI labs to prioritize safety and address potential risks from increasingly powerful models. But European competitors read it as a playbook for regulatory capture, one that would lock in advantages for companies that already sit at the frontier.

The accusation gained traction beyond Mistral’s own camp. Leading voices from Germany’s and France’s AI sectors echoed the framing, labeling Anthropic’s proposals as self-serving policies that threaten open innovation. Companies including Proton and Black Forest Labs have aligned themselves with the push to accelerate, not decelerate, European AI development.

Regulatory capture or legitimate risk management

Anthropic recently pushed for an antitrust waiver that would allow competing AI labs to collaborate on safety measures. European firms rejected this outright, arguing such waivers would create a cozy club of established players who get to set the rules of the road, with newcomers left navigating an obstacle course designed by their competitors.

The EU AI Act is already in effect, applying tiered obligations to AI systems based on risk levels. European policymakers are increasingly viewing US safety proposals through the lens of industrial strategy rather than genuine risk mitigation.

Mistral’s rapid ascent and the sovereignty play

Founded in 2023, Mistral secured a €3 billion Series D funding round that pushed its valuation to approximately €21 billion. The company has been actively securing government contracts within France, positioning itself as the domestic alternative to OpenAI, Google, and Anthropic for sensitive public-sector applications.

The emphasis on open-weight models is central to this vision. Unlike closed API-only services from major US providers, open-weight models allow governments and enterprises to run AI systems on their own infrastructure, inspect their inner workings, and modify them for specific use cases.