Anthropic AI export controls block foreign access to top models

Anthropic cut off access to Fable 5 and Mythos 5 for foreign users on Friday after the Trump administration restricted use of its most advanced AI models. The order also pushed the company to disable the systems for all customers while it seeks clearance from US officials.

By moving through export controls, the directive turns a product dispute into a harder regulatory test. Anthropic said the restriction applies to all foreign nationals, including foreign-national employees, and covers the two models it had just put at the center of its latest competitive push. Investors and enterprise customers now have to price in a less familiar risk: a top model tier that can lose part of its customer base by government order.

At 5:21 p.m. ET, Anthropic received the directive and suspended access while it works with officials, according to the company statement. Because the control was broad, the company said it had to disable Fable 5 and Mythos 5 for every customer, rather than only for accounts outside the US.

Without challenging the administration directly, Anthropic disputed the premise. “We believe this is a misunderstanding and are working to restore access as soon as possible,” the company said. That leaves several paths open: a quick license, a narrower directive or a longer policy review.

Axios reported that Commerce Secretary Howard Lutnick sent a letter to Anthropic chief executive Dario Amodei requiring a license for export, re-export or domestic transfer of the models. Under the process described by Axios, Anthropic would have to seek individually validated licenses, putting model access inside case-by-case federal approvals rather than a normal software sales channel.

National-security perimeter widens

That creates a plain commercial problem.

Once a model is classed as too sensitive for foreign governments, companies, individuals and some employees, the regulatory perimeter is no longer limited to chips, data centers or cloud capacity. It reaches into model weights, hosted services and the staff allowed to work with them. For an AI lab selling frontier capability as an enterprise platform, that is a different kind of operating risk.

Timing adds to the pressure on Anthropic. The Fable 5 and Mythos 5 launch had already put the company under scrutiny over how powerful models should be marketed, tested and limited. With the export-control order, the fight shifts from product governance to national security. A customer weighing a long-term contract now has to ask whether access to the top tier could turn on nationality, license status or a future Commerce Department interpretation.

Inside Washington, the step shows how quickly AI policy can move from voluntary safety commitments to hard access controls. Axios said the administration framed the models as national-security assets, not simply commercial software. Officials worried about frontier systems helping foreign rivals may welcome that argument, but it creates a sharper compliance burden for companies built around global teams and global customers.

What happens next depends on whether officials accept Anthropic’s misunderstanding claim. If they do, access could return under revised license terms and leave the episode as a warning shot. If they do not, Fable 5 and Mythos 5 become an early test case for model-level export controls, with investors left to judge how much of the frontier AI market remains commercially open once national-security agencies start drawing the boundaries.