Anthropic Suspends Customer Access to Fable 5 and Mythos 5 After U.S. Government Order
Miles Bennett
Anthropic said Friday it will immediately cut off all customer access to Fable 5 and Mythos 5 after receiving an export-control directive from the Trump administration — just three days after the models' public launch. It marks the first time the U.S. government has forced the recall of a commercially deployed AI model over jailbreak risk.
What exactly happened?
The Trump administration issued an export-control directive ordering Anthropic to immediately stop all customer access to Fable 5 and Mythos 5.
The order covers all foreign nationals inside and outside the U.S., including Anthropic's own foreign-national employees.
This means → it is not a voluntary product withdrawal but a government mandate. Other Anthropic models remain available, so the directive targets these two new models specifically.
Why did the government act so suddenly?
Axios reported the letter came from Commerce Secretary Howard Lutnick, triggered after another company claimed it could jailbreak Mythos.
In plain terms = someone showed government officials a way past the model's safety guardrails, and officials treated that as a national-security threat serious enough to order an immediate shutdown.
The Commerce Department oversees export controls — its direct intervention in AI model access is itself a signal that regulatory reach is expanding.
How did Anthropic respond?
Anthropic characterized the government's concern as a "minor vulnerability", saying the demo uncovered a small number of previously known, low-severity flaws.
The company stressed that other publicly available models can surface the same vulnerabilities without any bypass technique at all.
This means → Anthropic's subtext is clear: if this standard holds, every frontier model on the market should be recalled, not just Fable.
Are Fable's safety guardrails actually robust?
Anthropic said it conducted thousands of hours of red-team testing before Fable's launch, working with the U.S. government, the UK's AISI, and multiple private third parties.
The conclusion: Fable's guardrails are significantly stronger than any previously deployed model's.
But the company also acknowledged that no model provider can achieve perfect jailbreak resistance today. In plain terms = absolute security does not exist under current technology — the real question is where to draw the line.
What does this really mean going forward?
Anthropic explicitly disagreed with the government's logic: a "narrow potential jailbreak" should not justify recalling a model already deployed to hundreds of millions of users.
Yet the company is complying with the legal order — protest aside, the models are already offline.
This reflects a deeper tension: government control over frontier AI models is expanding fast, and the industry has no equal counter-mechanism. Once "jailbreak risk" becomes a recall standard, every new model launch faces similar political uncertainty.
Content is for reference only, not financial advice.