Anthropic CEO Calls for FAA-Style AI Testing While Launching Claude Mythos 5

On Wednesday, Anthropic CEO Dario Amodei released a policy essay calling for binding safety requirements and mandatory third-party testing for frontier AI models, modeled after FAA regulations for aircraft. The proposal mandates testing across four risk categories: cybersecurity, bioweapons, loss of AI control, and automated R&D, with government authority to block unsafe deployments.

The policy announcement coincides with Anthropic's launch of Claude Mythos 5, a restricted frontier model for cybersecurity organizations and government partners. The company simultaneously released Claude Fable 5, a public-facing version that routes sensitive requests to a less capable model as a safeguard against misuse.