Statement on the US government directive to suspend access to Fable 5 and Mythos 5

Details

In an unprecedented regulatory action, the US government invoked national security authorities to order Anthropic to suspend all access to Claude Fable 5 and Mythos 5 for any foreign national, whether inside or outside the United States — including Anthropic's own foreign national employees. The practical consequence: Anthropic disabled both models for its entire customer base to ensure compliance. The directive arrived at 5:21pm ET with no advance notice and provided no specific details of its national security concern.\n\nThe government stated it believes it became aware of a method for bypassing Fable 5's safeguards. Anthropic reviewed the demonstrated technique and found it exploited a small number of 'previously known, minor vulnerabilities' that are also discoverable through other publicly available models — a direct dispute of the government's threat assessment. Anthropic has publicly maintained that perfect jailbreak resistance is impossible for any model provider today, stating this clearly at Fable's launch.\n\nAnthropicmdefended its defense-in-depth strategy: safeguards were designed to make jailbreaks narrow or expensive, combined with monitoring for rapid response. The company required 30-day customer data retention to enable jailbreak detection and mitigation — a policy with real commercial costs that it accepted as a security measure.\n\nAll other Anthropic models (Opus, Sonnet, Haiku) remain available. Only Fable 5 and Mythos 5 were targeted. The action sets a direct precedent for government intervention in AI model deployment using existing export control legal authority.