Anthropic says it will remove access to two AI models, citing Trump administration directive on foreign nationals

Anthropic said it will remove access to two of its AI models, Fable 5 and Mythos 5, after the company said it received a Trump administration directive requiring additional restrictions tied to user nationality. The company announced the decision on Friday, describing it as necessary to comply with the government request while it works to ensure continued access to other Anthropic models.

In its statement reported by The Hill, Anthropic said it plans to take the two models down rather than keep them available under the existing access scheme. The company did not make public details of the government request, including how long the restrictions are expected to last or what specific compliance conditions the administration imposed.

The Hill reported that the length and other details of the directive have not been made public, and Anthropic did not provide further documentation of the request in the account cited by the outlet. The company’s explanation, according to The Hill, centered on the administration’s stated security concerns and a requirement to limit access for foreign nationals.

The decision comes as the U.S. government has increasingly emphasized national security and risk management for advanced AI systems. Restricting access by nationality can change the compliance obligations for AI developers and for platforms that provide model access, because the operator must determine who is eligible to use restricted systems and how to enforce those limits.

For Anthropic, removing access to two specific models also affects customers that rely on those versions for particular capabilities and workflows. Any sudden change in availability can require customers to switch to alternative models or wait for reinstatement once the underlying conditions are clarified or expire.

The administration directive raises questions about process and due process for regulated technology providers, particularly when the underlying legal authority and scope are not publicly described. Because Anthropic and the reporting outlet did not disclose the directive’s duration or legal framework, it remains unclear what internal standards will govern future changes or whether the restrictions could be modified.

Anthropic indicated that it is taking steps to comply with the directive, but it has not provided additional public information that would allow outside parties to assess the precise compliance boundaries. The next development will likely depend on what the administration and Anthropic disclose about the directive’s duration, enforcement mechanism, and whether access to the removed models will return.

The Hill’s report also underscores that the administration’s request, as described by Anthropic, did not involve an immediate public release of the directive text or the full set of conditions. Until more information is available, customers and researchers will have to rely on Anthropic’s announcements to understand which models remain accessible and under what rules.