Anthropic has offered the European Union access to its Mythos artificial intelligence model, marking the first time this advanced cybersecurity software will be available outside of the US and UK. A spokesperson from the EU’s cybersecurity agency Enisa confirmed ongoing negotiations regarding the conditions of this access. This development follows meetings between European Commission officials and Anthropic in San Francisco, where discussions focused on potential participation in Project Glasswing, aimed at addressing security vulnerabilities. Importantly, Anthropic has been working to develop stronger cyber safeguards to ensure the safe deployment of its models, which have previously been limited to select entities, including the UK government, that evaluated Mythos prior to its rollout.

Anthropic: Anthropic is an AI company developing advanced models focused on safety and capability, including specialized cyber security applications. It has offered the EU access to its Mythos model, marking the first expansion of this capability beyond the US and UK. The company is also progressing toward broader customer release while collaborating closely with the US government on national security uses.
Thomas Regnier: Thomas Regnier serves as a spokesperson for the European Commission, providing updates on EU-Anthropic negotiations. He confirmed productive meetings in San Francisco and emphasized the importance of the Mythos access for assessing AI-related risks. His role highlights ongoing bilateral cooperation between the Commission and the AI firm.
European Commission: The European Commission is the executive branch of the European Union tasked with policy implementation and international engagement on technology matters. It has pursued access to Anthropic’s Mythos model through direct talks and visits to company headquarters, while exploring participation in the Project Glasswing industry coalition. The effort aims to enhance the bloc’s understanding of advanced AI cyber capabilities.

Model Expansion: Anthropic is developing stronger cyber safeguards to enable release of Mythos-class models to all customers.
Cybersecurity Access: The EU’s cyber security agency Enisa is in talks to use the Mythos model under conditions still being negotiated.
Allied Government Access: The UK government has gained access to Mythos through its AI Security Institute.