The company said the practical effect of the directive was that it had to disable Fable 5 and Mythos 5 for all customers to ensure compliance. Anthropic said access to its other models was not affected.
The direct government document was not found on the Bureau of Industry and Security homepage or the Federal Register's Bureau of Industry and Security agency page during drafting. The legal basis and any licence condition should therefore be described as a directive reported and described by Anthropic, not as a published Federal Register rule unless Commerce publishes one.
Anthropic said it received the directive at 5:21pm Eastern time on June 12. The company said the government letter did not provide specific details of the national-security concern. Anthropic said its understanding was that the government had become aware of a method of bypassing, or "jailbreaking", Fable 5, and that Anthropic had reviewed a demonstration involving a small number of previously known, minor vulnerabilities.
The models had launched three days earlier. In its June 9 launch post, Anthropic described Fable 5 as a Mythos-class model made available for general use and Mythos 5 as the same underlying model with safeguards lifted in some areas for selected cyber defenders and infrastructure providers. Anthropic said Mythos 5 would initially be deployed through Project Glasswing in collaboration with the US government.
The technical dispute is over the safety threshold for deployed models. Anthropic said it had used safeguards that route some cybersecurity, biology and chemistry, and distillation-related requests away from Fable 5 to Claude Opus 4.8. It also said more than 95% of Fable sessions involved no fallback in early data, and that external testing had not found a universal jailbreak in more than 1,000 hours of bug-bounty testing.
