According to Cointelegraph, AI safety firm Anthropic announced it will open access to its AI model Mythos to the European Union Agency for Cybersecurity (ENISA) to support the agency's work in assessing AI system safety. The collaboration is aimed at assisting EU regulators in evaluating AI risks and developing corresponding standards.
Collaboration Details and Objectives
Anthropic, a research company deeply focused on AI safety, has developed Mythos with advanced capabilities in language understanding and generation. ENISA, the EU's dedicated cybersecurity authority, has increasingly integrated AI safety into its regulatory framework in recent years. Through this partnership, ENISA will be able to directly analyse Mythos' technical architecture and verify the effectiveness of its safety mechanisms.
By granting model access, ENISA can base its recommendations on practical technical evaluations, providing a scientific foundation for EU AI regulatory policy. This move not only aids in harmonising internal EU standards, but may also positively influence the global AI safety assessment framework, promoting the establishment of transparent and verifiable evaluation mechanisms.