Trump administration forces Anthropic to pull Fable 5 and Mythos 5 in first federal AI shutdown

A Commerce Department export-control directive citing a foreign-national access risk took both models offline within hours of arriving Friday.

Anthropic took Claude Fable 5 and Claude Mythos 5 offline late Friday, hours after a Commerce Department export-control letter landed at 5:21 p.m. Eastern citing a foreign-national access risk. Per NBC News, it’s the first time the federal government has forced a leading U.S. AI company to retract systems from public use.

Fable 5, built on Mythos with additional restrictions, had been announced only the previous Tuesday and described in Anthropic’s blog post as “safe for general use.” The company says the directive “did not provide specific details,” and that officials verbally cited a jailbreak technique Anthropic characterized as one that “appear[ed] relatively simple” and replicable on other public models, including OpenAI’s ChatGPT.

The political backdrop is louder than the technical one. On June 10, two days before the letter, CEO Dario Amodei published a blog post calling Mythos an “emblematic example” of frontier-model national-security risk and demanding mandatory third-party testing. The Wall Street Journal, cited by Tech Policy Press, reports Amazon CEO Andy Jassy was the channel that instigated the Friday letter. Anthropic had filed confidentially for an IPO earlier this month at a $965 billion valuation. This is the second time it’s been blacklisted by the federal government.

Former AI and crypto czar David Sacks has publicly branded the company “woke,” “leftist,” and accused it of running “a sophisticated regulatory capture strategy based on fear-mongering.” On Monday the Department of War posted on X that it has moved at least two-thirds of its AI workflows off Anthropic’s models and “will no longer be single-threaded to one AI provider.”

Senior Anthropic staff spent the weekend in Washington arguing the directive down. The structural read is harder to miss: a lab that spent a year lobbying for a federal safety regime got one, aimed at itself, written by people who don’t like it.