Anthropic has suspended its state-of-the-art artificial intelligence models after the US government directed the company to limit access for all foreign nationals, citing national security concerns.
Anthropic received the directive from commerce secretary Howard Lutnick on Friday afternoon, after the government discovered a “jailbreak”, or a way of bypassing safety guardrails built into the model. Anthropic added that the “net effect” of the order, confirmed by a US official, was that it was forced to switch off access everywhere, including in the US.
The Trump administration has had access to Anthropic’s frontier Mythos model for several weeks, and several officials had raised concerns about its potential use to exploit cybersecurity vulnerabilities in critical infrastructure, such as the global financial system. Federal agencies were experimenting with the model ahead of its release.
[ The AI bubble question: is Anthropic justified at $1tn?Opens in new window ]
However, the administration had not prevented Anthropic from deploying Mythos to its customers this week. President Donald Trump recently signed an executive order that stopped short of giving the government powers to block the release of a model, even as it set up a voluntary framework for various agencies to inspect frontier models in advance of their wider roll-out.
RM Block
Anthropic said on Friday it had withdrawn its most powerful models, Mythos 5 and Fable 5, from all users. But the company, which is locked in a legal battle with the US government, condemned the order and said it was based on a “misunderstanding”.
The disagreement is likely to inflame a relationship which has become increasingly fractious over the course of the year as Anthropic and members of the government have sparred over how and when to deploy the powerful technology.
“We believe the government should have the ability to block unsafe deployments, as part of a statutory process that is transparent, fair, clear and grounded in technical facts. This action does not adhere to those principles,” Anthropic said in a statement.
“We disagree that the finding of a narrow potential jailbreak should be cause for recalling a commercial model deployed to hundreds of millions of people. If this standard was applied across the industry, we believe it would essentially halt all new model deployments for all frontier model providers,” it added.
Anthropic is suing the US government over its designation by the Pentagon as a “supply-chain risk” following a dispute over the use of its models in warfare. It has faced criticism for its safety policies from various members of Trump’s coalition, including former AI tsar David Sacks. Trump himself has said the company is run by “left-wing nut jobs”.
The president has also recently suggested AI companies should donate equity to the government, although that plan has yet to get off the ground.
The White House did not immediately respond to a request for comment. – Copyright The Financial Times Limited 2026
