At 5:21 p.m. Eastern on Friday, an order from the Commerce Department arrived at Anthropic instructing the company to disable Claude Fable 5 and Mythos 5, its two most capable systems, for every customer in the world. By late Friday night, both models were offline. It’s, by any reasonable accounting, the first time the United States government has compelled a frontier AI laboratory to take a deployed model down.

The directive was drafted by Commerce Secretary Howard Lutnick with officials at the Bureau of Industry and Security, and framed as an export-control action barring any foreign national, inside or outside the country, from using the systems. A U.S. official confirmed the letter to Bloomberg. The landing pages went dark with a single line: “Fable 5 is temporarily unavailable.”

What the order doesn’t contain, according to Anthropic, is a specific reason. The company says it believes the action stems from a claimed method of “jailbreaking” Fable 5, demonstrated to the government with only verbal evidence so far. Anthropic’s own review of the technique characterized it as “a small number of previously known, minor vulnerabilities,” and noted that the underlying capability is already present in other publicly available systems, including OpenAI’s GPT-5.5.

The timing is its own story. Fable 5 had launched three days earlier as the general-availability descendant of Mythos, which until Friday had been restricted to roughly 50 vetted organizations under a program called Project Glasswing. Amazon, Apple, Google, Microsoft and CrowdStrike were among the participants. The Financial Times reported last week that the National Security Agency had been using Mythos for offensive cyber operations. The same week the government appears to have decided the broader release was intolerable, its own agencies were operational customers of the restricted tier.

Anthropic’s posture is unusually pointed for a company that has built its brand on cooperating with regulators on safety. “We believe the government should have the ability to block unsafe deployments, as part of a statutory process that is transparent, fair, clear, and grounded in technical facts. This action does not adhere to those principles,” the company said. The statement lands inside an existing fight: earlier this year the Department of Defense designated Anthropic a supply-chain risk, and the company sued the Trump administration to reverse the finding. That litigation is ongoing.

The structural read is harder to miss. A lab that spent years arguing the government should’ve authority to halt unsafe models has now had that authority used against it, on grounds it considers thin, with reasoning delivered orally. The safety-pilled position and the export-controlled position turn out to be the same position, only the chair has changed.

Sources