Anthropic suspends access to its most advanced AI models after U.S. government directive

Anthropic said Friday that it is cutting off access to its two most advanced models, Claude Fable 5 and Claude Mythos 5, after receiving a U.S. government directive citing national security concerns. The move came only days after Fable 5 was launched for the general public.

According to the company, the U.S. Commerce Department ordered it to restrict the models’ use by “any foreign national, whether inside the United States or outside it, including Anthropic employees who are foreign citizens.” The concern, which Anthropic said was not fully detailed to it, apparently involves the possibility of bypassing some of the models’ safeguards, potentially enabling software vulnerability discovery and advanced cyber operations.

Rather than block only foreign users, Anthropic chose to suspend both models for all users temporarily in order to ensure full compliance. The company noted that Fable 5 was the first model from the Mythos family made available to the public, and that access to Mythos models had until recently been limited to a small number of preapproved organizations because of their advanced cyber capabilities.

Anthropic also criticized the government action, saying regulators should be able to stop unsafe deployments through “a transparent, fair, clear statutory process based on technical facts,” and argued this case does not meet those standards. The company is already on the Pentagon’s blacklist, which says its systems are too dangerous for government use, and Anthropic has sued the Trump administration over that designation. The decision may signal a broader shift in U.S. AI policy, as export restrictions have so far focused mainly on chips and advanced hardware, but now are aimed directly at AI models themselves.