EU Cybersecurity Agency Gets Access to Test Anthropic and OpenAI's Latest AI Models

A European Commission spokesperson confirmed that the EU Agency for Cybersecurity (ENISA) has been given access to test Anthropic's Mythos 5 AI model. The spokesperson also said ENISA was granted access to OpenAI's latest model, GPT-6-Astra.

Previously, the White House introduced restrictions barring foreign entities from accessing high-performance AI models from US companies. Although those limits were later eased, it remained unclear whether foreign agencies could gain access.

The EU agency's testing access signals a new shift in the US policy balance between AI model exports and security reviews.