Anthropic is moving to provide its cybersecurity AI model, Mythos, to the European Union (EU). This would mark the first time the technology has been made available outside the United States and the United Kingdom.
On June 1 (local time), CNBC and other outlets reported that the EU and Anthropic are negotiating detailed terms for access to Mythos.
The European Union Agency for Cybersecurity (ENISA) confirmed it is discussing the adoption of Mythos and said officials have yet to finalize agreements on access rights and operating procedures.
The European Commission recently sent officials to San Francisco to negotiate with Anthropic. The EU is seeking to participate in Anthropic’s cybersecurity collaboration, Project Glasswing, to help identify vulnerabilities in Europe’s critical systems and strengthen defensive capabilities.
Toma Renier, a spokesperson for the European Commission, said, “We have held several productive meetings with Anthropic and welcome recent progress toward securing potential access. This is important for clarifying the risks that AI technologies can pose.”
So far, Anthropic has made Mythos available only to a limited group of U.S. companies and institutions through Project Glasswing, including Microsoft, Apple, JPMorgan Chase and CrowdStrike.
Anthropic has also worked closely with the U.S. government to test and deploy Mythos for national security purposes and for sensitive government agencies. Outside the U.S., only the U.K. government has evaluated the model and gained access through the AI Security Institute (UK AISI).
The prospect of EU access is notable because the U.S. government has been cautious about sharing advanced AI models overseas. The EU has sought access to Mythos for months and has recently discussed the issue with U.S. officials.
In May, the EU secured access to OpenAI’s “GPT-5.5-Cyber” model, but experts say Mythos is assessed to have stronger cybersecurity capabilities. European authorities have emphasized the need to evaluate directly both the risks and practical uses of next-generation AI security tools.
Anthropic CEO Dario Amodei has said, “I hope that the U.S. and allied governments will use this technology to protect democracy and security,” and he warned that it must not be used for authoritarian purposes or citizen surveillance.
Anthropic also announced plans to make a model with Mythos-level performance available to a broader set of customers soon. “Models at this level require stronger safeguards before broad release,” the company said, adding that it is rapidly developing those protections and expects to offer Mythos-level models to more customers within weeks.
By Chan Park (cpark@aitimes.com)
Most Commented