According to Anthropic in a post, the directive from U.S. authorities applies to all foreign nationals, whether inside or outside the United States, including employees of the company who are not U.S. citizens. Access to the company’s other Claude models is not affected, as the company clarified.

According to the New York Times, citing a person familiar with the matter, the restriction order on the new AI models came from the U.S. Department of Commerce. It remains unclear how long the restrictions will last, while Anthropic did not comment beyond its public statement, and the White House did not immediately respond to a request for comment.

The order is said to have been issued under export control regulations, moving the AI debate into a new phase: from restrictions on advanced chips and infrastructure to restrictions on access to AI models themselves. Axios reported that the Trump administration imposed strict limits on Anthropic’s two most powerful models, treating them as technologies with potential national security implications. The decision affects governments, companies, and even users.

Mythos 5 had already attracted significant interest but also concern in Washington and Silicon Valley, as it is associated with advanced capabilities in cybersecurity. Anthropic had presented Fable 5 as a more limited version of Mythos 5, with additional safety filters for sensitive domains such as cybersecurity, biology, and other high-risk areas.

The company, however, strongly hinted that it considers the government decision excessive. “We apologize for the disruption to our customers. We believe this is a misunderstanding and are working to restore access as soon as possible,” it said in its statement.

The decision could have immediate consequences both for the commercial use of the new models and for internal technological development at Anthropic, since the restriction also applies to foreign nationals working at the company. This is a much broader move than typical export controls on advanced technologies, as it does not only restrict product availability to specific countries but access based on user nationality.

The blockade comes at a time when the U.S. government is seeking a new framework for overseeing the most advanced AI models, after months of intense debate about whether systems with high cybersecurity capabilities could be used not only for defense but also for attacks on critical networks. Experts disagree on whether such technologies should be strictly restricted, as the same tools can be used both by hackers and by organizations trying to detect and fix vulnerabilities before they are exploited.

The case now opens a new front in Washington’s relationship with major AI companies. Until now, U.S. policy has focused mainly on controlling advanced semiconductors and access to computing power by countries such as China. With the Anthropic case, however, the U.S. government appears to be intervening directly in the use of the models themselves, whenever it considers their capabilities to touch the core of national security.