Anthropic opens Mythos to EU after months of cybersecurity talks

Anthropic has agreed to give European Union institutions access to its most advanced AI model, Mythos, following sustained pressure from Brussels over cybersecurity risks. The European Commission confirmed it held 'several productive meetings' with the U.S. company, with tech sovereignty spokesperson Thomas Regnier stating the bloc wants to better understand the technology's potential dangers. Mythos originally launched in April under Project Glasswing, Anthropic's cybersecurity initiative, and was limited to a select group of companies. The model proved highly effective at detecting software vulnerabilities, but its release sparked widespread concern among governments, banks, and tech firms about misuse by malicious actors. 'Let's not forget that Mythos is not a one-off; a new wave of powerful models are coming to market,' Regnier said. 'This is a shared challenge, and we are intensifying our discussions with like-minded partners, including the United States.' Official terms remain unclear. Anthropic had earlier told the Commission it needed U.S. government approval, and sources indicated Washington was reluctant to share the model broadly to maintain its AI leadership. The EU previously secured access to OpenAI's GPT-5.5-Cyber in May, but Commission officials noted talks with Anthropic were at a 'different stage.' The Trump administration separately reached agreements in May with Google DeepMind, Microsoft, and Elon Musk's xAI to review AI models before public release. The Mythos deal signals a shift in transatlantic cooperation on high-risk AI, though the exact scope of access and safeguards is still being finalized.