At 5:21 p.m. ET on Friday, a Commerce Department letter signed by Secretary Howard Lutnick landed in Dario Amodei’s inbox, and within hours Anthropic had disabled Claude Fable 5 and Mythos 5 worldwide. The export-control directive named both models specifically, citing national security, and because it covered foreign nationals inside and outside the United States, including Anthropic’s own staff, the company concluded it had no clean way to enforce the order short of a global shutdown.

Fable 5 had been public for three days.

Anthropic’s statement is careful but unmistakably aggrieved. The jailbreak that triggered the action, the company says, surfaced only “previously known, minor vulnerabilities” that other publicly available models including OpenAI’s GPT-5.5 can also locate. Anthropic adds it has received only verbal notice of the technique. The models in question, it notes pointedly, had been “deployed to hundreds of millions of people.” The company says it’s “working to restore access as soon as possible.”

The structural irony is that Anthropic’s own marketing posture probably accelerated this. Mythos 5 was sold as so capable at finding software flaws that it had to be locked behind Project Glasswing, a restricted access tier of roughly 50 vetted partners including Amazon, Apple, Google, Microsoft, and CrowdStrike. That framing is catnip to a Commerce Department already hunting for export-control hooks on frontier AI. Sam Altman saw the trap coming. On Ashlee Vance’s podcast in April he called Anthropic’s posture “fear-based marketing,” a line that now reads less like a competitive jab and more like a forecast.

The timing compounds the damage. Anthropic is widely expected to IPO this year. SpaceX went public Friday at a $2.1 trillion market cap, resetting investor expectations for what a politically aligned frontier-tech listing can clear. Anthropic, meanwhile, is already suing the administration to reverse a Pentagon designation labeling it a supply-chain risk, a category CNBC notes is typically reserved for foreign adversaries.

The company spent years arguing its models were uniquely dangerous. The government finally agreed.

Sources