The US government on Friday ordered Anthropic to immediately shut down access to two of its most powerful artificial intelligence models – Claude Fable 5 and Claude Mythos 5 – citing national security concerns. Humane announced in X that he has complied, but made it clear that he believes that the government I got it wrong.
The directive, which Anthropic said it received Friday at 5:21 p.m. ET, forcing the company to disable both models for all users worldwide — not just foreign nationals, who were nominally targeted by the government’s export control order. Access to other Anthropic models is not affected.
Why does any of this matter? Mythos is Anthropic’s most capable AI model, one that the company previewed in early April and has since kept tightly constrained due to what Anthropic described as its exceptional ability to detect security vulnerabilities in software. According to Anthropic, Mythos found flaws in every major operating system and web browser it tested, so instead of releasing it widely, the company launched a controlled program called Project Glasswing, which it shared with about 50 vetted organizations, including Amazon, Apple, Google, Microsoft and CrowdStrike, to use for defensive cyber work.
Fable 5, released just three days ago, was Anthropic’s response to apparent commercial pressure: a version of Mythos equipped with guardrails that block responses in high-risk areas such as cybersecurity and biology, making it safe enough for general release, the company argued. It was immediately the most capable AI model publicly available, according to benchmark tests by Vals AI, a company that tracks AI technology performance.
The government directive is framed as an export control action, limiting foreign nationals’ access to the models. But in one long blog postAnthropic says it understands the underlying concern is an alleged jailbreak of Fable 5. So far, the company says, the government has only provided verbal evidence of a “potential narrow, non-universal jailbreak” — which, as Anthropic describes it, amounts to forcing the model to read a specific codebase and software. And by the way, the company adds, it’s a “capability level” that’s already widely available in other publicly accessible models, including OpenAI’s GPT-5.5. It’s also regularly used by cybersecurity professionals for defensive purposes, Anthropic says.
Anthropic’s broader argument is that its strongest safeguards work through independent classifier systems that operate separately from the model itself, meaning that even if someone convinces Fable to keep talking beyond a denial, the underlying protections against more dangerous outcomes remain in place.
Clearly, none of this was enough to stop the government from acting, and Anthropic is not hiding its disappointment. “We disagree that the finding of a limited potential jailbreak should be grounds for recalling a commercial model that has been deployed to hundreds of millions of people,” the company wrote. “If this standard were implemented across the industry, we believe it would essentially stop all new model developments for all frontier model providers.”
Anthropic is widely expected to seek an IPO this year and has staked much of its public identity on being the savvy alternative to rivals. The irony is not lost on observers that the very care Anthropic took in curtailing Mythos — which it promoted as a model so dangerous it couldn’t go public — has now attracted exactly the kind of government scrutiny that could most disrupt its operations.
OpenAI’s Sam Altman must be enjoying this, at least. In April, he told podcaster Ashlee Vance that Anthropic’s handling of Mythos amounted to “fear-based marketing.” “It’s clearly incredible marketing to say, ‘We’ve built a bomb.’ We were about to drop it on your head. We’ll sell you a bomb shelter for $100 million,” Altman said. Altman, whose company is also widely expected to seek an IPO as soon as possible, didn’t predict a government shutdown, but he identified something that has come back to Anthropic for now, which is that when you spend months telling the world your AI is uniquely dangerous, the world — including the government of the US – tends to listen.
When you purchase through links in our articles, we may earn a small commission. This does not affect our editorial independence.
