Imagine you are preparing to use a new AI model, and access disappears overnight. That is what Anthropic said happened after the U.S. government issued an export control directive affecting Claude Fable 5 and Claude Mythos 5. The company said the directive cited national security authorities and required suspension of access by any foreign national, whether inside or outside the United States. Anthropic said it disabled both models for all customers to ensure compliance. The action matters because it shows how advanced AI systems can move from product launch to government restriction in a very short period.
The immediate parties are Anthropic, its customers, foreign national users, and U.S. authorities. Anthropic said the order also covered foreign national Anthropic employees. Fable 5 had been presented as a broadly available model with safeguards, while Mythos 5 was described as the same underlying model with some safeguards lifted for a smaller set of vetted partners. People who should care include developers, enterprise buyers, cybersecurity teams, life science researchers, and policy officials who depend on clear rules for access to frontier AI systems.
The shutdown followed Anthropic’s June 9, 2026, announcement of Claude Fable 5 and Claude Mythos 5. In that launch post, Anthropic said Fable 5 was available broadly, while Mythos 5 was restricted to Glasswing partners and planned trusted-access programs. Anthropic said Fable 5 and Mythos 5 were priced at $10 per million input tokens and $50 per million output tokens. On June 12, 2026, Anthropic said it received the government directive at 5:21 p.m. ET and that access to all other Anthropic models would not be affected. Public sources do not clearly confirm the government’s full written rationale beyond Anthropic’s statement that the letter did not provide specific details of the national security concern.
In practice, the issue turned on access control and safeguards. Anthropic said Fable 5 used safeguards so that some sensitive cybersecurity, biology, and chemistry queries would be routed to Claude Opus 4.8. The company said Mythos 5 was intended for a small group of cyberdefenders and infrastructure providers through Project Glasswing, with some cyber safeguards lifted. Anthropic said its understanding was that the government had become aware of a possible method of bypassing, or jailbreaking, Fable 5. It also said the reviewed demonstration involved finding a small number of previously known, minor software vulnerabilities. A useful analogy is a locked laboratory door: the public may see the building, but only approved people are allowed into the sensitive rooms.
What comes next is uncertain. Anthropic said it disagreed that a narrow potential jailbreak should justify recalling a commercial model, while also saying it was complying with the legal directive. The company said it was working to restore access as soon as possible, but public sources do not clearly confirm when or whether that will happen. The practical next step for affected users is to check Anthropic’s official availability or status information before planning work around Fable 5 or Mythos 5. More broadly, a general best practice is to maintain fallback models and clear vendor-risk plans when relying on newly released AI systems.