Anthropic is granting the European Union access to its AI model Mythos, following months of discussions prompted by the EU's concerns over cybersecurity risks associated with the technology. The European Commission confirmed productive meetings with Anthropic, indicating a desire to understand the potential threats posed by AI advancements.
Mythos, which was initially launched in April as part of Anthropic's Project Glasswing, is designed to identify security vulnerabilities in software, raising alarms about its potential misuse by malicious actors. EU spokesperson Thomas Regnier emphasized the need for collaboration with partners like the U.S. to address shared challenges in AI technology.
While the specific terms of the access agreement remain unclear, the U.S. government has expressed reluctance to share powerful AI models with foreign governments, aiming to maintain its leadership in the AI sector.
This situation follows the EU's earlier access to OpenAI's GPT-5.5-Cyber model and reflects a broader trend of governments seeking to assess and regulate AI technologies before their public release