Anthropic has agreed to provide the European Union with access to its most advanced AI model, Mythos. The CNBC reported the development, and the European Commission later confirmed it had held "several successful consultations" with Anthropic.
Why the access matters
Thomas Regnier, spokesperson for the European Commission on technology sovereignty, welcomed the news and said the EU seeks a fuller understanding of the technology’s potential risks. Mythos is noted for its ability to discover software security flaws and vulnerabilities, and its capabilities have alarmed governments, banks and tech companies — reporting indicates the model can identify thousands of previously unknown vulnerabilities.
Background: Project Glasswing and the limited release
Anthropic initially made Mythos available in April on a limited basis under a cybersecurity program called Project Glasswing. Because of the model’s capabilities, the EU pursued months-long access negotiations, citing cybersecurity concerns.
Transatlantic consultations and U.S. clearance
Anthropic informed the Commission that sharing access with the EU required authorization from the U.S. government. In recent weeks the European Commission accelerated talks with the United States; reports say Washington initially opposed sharing the model with foreign governments, as the U.S. seeks to maintain a leading position in artificial intelligence.
Earlier, in May, the EU gained access to OpenAI’s GPT-5.5-Cyber model. At that time, negotiations with Anthropic were still in early stages — the Commission said only four to five consultations had taken place then.
Outstanding details and broader challenges
The precise terms of the current agreement have not been disclosed. Regnier also noted that Mythos is not an isolated case: a new wave of high-capability AI models is entering the market, posing common challenges for regulators and policymakers. The EU is therefore maintaining closer dialogue with like-minded partners, including the United States.
Summary
The agreement between Anthropic and the European Commission enables the EU to perform security and risk assessments of the Mythos model. Key details about the sharing arrangements remain private, and transatlantic approvals played a significant role in enabling access.
