Anthropic's Most Powerful Model Pulled Offline by US Export Control Order

Anthropic's most capable AI model, referred to as Fable 5, has been taken offline globally since June 12 following an emergency directive from the US Commerce Department citing national security concerns. The shutdown was triggered after a method to bypass the model's safeguards was discovered, and Anthropic was unable to verify the geographic origin of requests in real time. Unlike typical rate-limiting or deprecation, the model is export-controlled, making it unavailable even to US users, with only a limited exception granted to select cyber-defense organizations. The simultaneous release of Sonnet 5, a cheaper and more capable agentic model, has drawn most public attention, overshadowing the significance of the policy-driven shutdown. The episode highlights a new risk for AI-dependent businesses: frontier model access can now be revoked by government order, making vendor diversification and system reliability more critical than chasing peak model performance.