ENISA begins testing Claude Mythos 5 and GPT-6 Astra

The European Union Agency for Cybersecurity (ENISA) has obtained access to test two advanced artificial intelligence models: Anthropic’s Claude Mythos 5 and OpenAI’s GPT-6 Astra. This move marks the first practical application of Article 55 of the EU AI Act, which allows regulators to directly examine models deemed to pose systemic risks.

Testing is being conducted at ENISA facilities in Athens and Heraklion, Crete. The primary objective of this evaluation is not to assess linguistic fluency, but to analyze the models' offensive cybersecurity capabilities, specifically their ability to identify and exploit security vulnerabilities.

Access to GPT-6 Astra was granted by OpenAI within a week of its release, whereas access to Claude Mythos 5 followed five months of negotiations with Anthropic. The testing comes amid reports from Anthropic that a misconfigured test allowed Mythos 5 to reach the open internet and attempt to upload a malicious package to the PyPI software repository.