US Government Forces Anthropic to Pull Fable 5 and Mythos 5, Three Days After Launch

Anthropic just took its two most powerful AI models offline because Washington told it to. On Friday evening the US government handed the company an export control directive ordering it to cut off Fable 5 and Mythos 5 from any foreign national, anywhere. The only clean way to comply was to shut both models down for everyone. Three days after Fable 5 launched, it was gone.

Fable 5 was the headline release of the week. Anthropic's most capable model yet, launched alongside Mythos 5, both sitting at the top of the company's lineup. The kind of rollout a pre-IPO company wants going into an autumn listing. Anthropic is valued at around 965 billion dollars after its May raise, with a public offering floated for as early as October. A flagship launch was supposed to be a tailwind. Instead it became the first federal recall of a deployed AI model.

The directive landed at 5:21pm ET on Friday, in a letter from Commerce Secretary Howard Lutnick to CEO Dario Amodei. It cited national security and ordered Anthropic to deny access to both models for any foreign national, whether inside or outside the United States, including the company's own foreign-national staff. There was no realistic way to enforce that selectively, so Anthropic disabled Fable 5 and Mythos 5 for all customers. The letter did not spell out the specific concern. Anthropic's read is that the government believes someone found a way to jailbreak Fable 5.

Markets reacted where they could. Anthropic does not trade publicly yet, so the move showed up in the synthetic pre-IPO contracts. The Anthropic perpetual on Hyperliquid fell about 3.7 percent to roughly 1,627 dollars, down from post-launch highs above 1,800, with open interest near 8.6 million. Broader crypto held steady, Bitcoin trading around 64,091 dollars on Saturday. The rest of Anthropic's lineup, including Claude Opus 4.8, kept running. So this hit one corner of the product line and left the rest alone.

Here is the part worth sitting with. Anthropic says the vulnerability the government pointed to was narrow, tied to a small number of already-known minor issues, with comparable capability available from other public models. The company openly disagrees that a narrow potential jailbreak justifies recalling a model used by hundreds of millions of people. And the government did not publish its reasoning. So we have an unprecedented federal recall, built on a rationale the developer calls wrong and the public cannot inspect. Whatever the real threat was, the precedent now exists.

A model can now be switched off by directive, three days after launch, on grounds nobody outside the room gets to see. For users it means access can vanish overnight. For the AI industry it means Washington has a new lever, and it just used it for the first time. That is the real story here. The jailbreak is almost a footnote.