Particle.news

U.S. Order Forces Anthropic to Disable Its Two Most Powerful AI Models

The company says it shut Fable 5 and Mythos 5 worldwide after a Commerce directive citing a national‑security 'jailbreak' that it says it cannot reliably block by nationality and that raises questions about export controls for cloud AI.

Overview

  • Anthropic disabled Fable 5 and Mythos 5 for all users after a Commerce Department directive on Friday told the company to block access for foreign persons and entities.
  • The government cited an alleged method to bypass the models' safety limits, often called a 'jailbreak', as the national‑security reason for the restriction.
  • A security researcher reported a jailbreak of Fable 5 within 24 hours of release, but Anthropic says the flaws shown were limited, patchable, and not a universal bypass of protections.
  • Anthropic said it could not reliably filter users by nationality in cloud service settings, which forced a global shutdown of the two models while other Claude models remain available.
  • The move updates a months‑long clash with U.S. authorities that includes a Pentagon supply‑chain designation and ongoing litigation, and it could set new precedents for applying export controls and pre‑release government review to frontier AI.