EU cybersecurity agency gains access to Anthropic's top AI model Mythos 5
The European Union’s cybersecurity agency ENISA has secured permission to test Anthropic’s most advanced AI system, Mythos 5, after months of negotiations.
ENISA announced that it has finally obtained a testing licence for Mythos 5, the most powerful artificial-intelligence model developed by the US-based company Anthropic. The access follows several months of talks that started in June, during which the EU sought to evaluate the model amid rising cybersecurity worries. Anthropic limits the full model to a small group of partners, providing a toned-down version named Fable for public use.
The European agency will now conduct trials to assess potential risks and benefits. ENISA’s experience includes earlier evaluations of OpenAI’s cutting-edge systems, most recently GPT-6-ASTRA. The move reflects heightened alarm among engineers and experts about the possibility of uncontrollable AI behavior.
