EU Commission meets with Anthropic in San Francisco as AI access negotiations heat up

The European Commission sat down with Anthropic in San Francisco on Thursday, a meeting that underscores just how central AI diplomacy has become to transatlantic relations. The talks centered on European access to Anthropic’s Mythos AI model, a system reportedly capable of uncovering thousands of cybersecurity vulnerabilities.

European Commission spokesperson Thomas Regnier confirmed the meeting, describing the discussions as productive. Anthropic followed up by extending a formal invitation to the EU’s cyber agency, proposing controlled access to Mythos under stringent security protocols.

What’s actually being negotiated

The EU has been locked out of some of America’s most advanced AI tools due to US export controls on high-capability AI systems. European officials have characterized these restrictions as discriminatory against allied partners.

The Mythos model sits at the center of these tensions. Anthropic’s system has demonstrated the ability to identify thousands of cybersecurity vulnerabilities, the kind of capability that could dramatically strengthen Europe’s digital defenses.

The formal invitation from Anthropic to the EU’s cyber agency represents a notable shift. Previously, negotiations had stalled. The fact that Anthropic is now proposing a structured pathway for access, even with strict security conditions attached, suggests both sides see value in making this work.

The export control backdrop

The EU is simultaneously conducting its own review of US export controls to determine whether these regulations are unfairly blocking collaborations with allied nations.

US export controls on advanced AI have been a growing source of friction with allies, not just adversaries. The controls were originally designed to prevent cutting-edge AI technology from reaching countries like China and Russia. But the broad strokes of the regulations have caught friendly nations in the same net, creating awkward diplomatic situations exactly like this one.

The controlled access framework that Anthropic proposed, with stringent security protocols, appears designed to satisfy both US regulatory requirements and European demands for inclusion.

Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.