Anthropic cuts off Fable and Mythos after U.S. export order

Anthropic disabled access to its Fable 5 and Mythos 5 AI models late Friday after it said the U.S. Commerce Department used export controls to block the company from giving the systems to foreign nationals. The action affects one of the company’s newest AI releases and raises new questions about how Washington may restrict powerful AI systems on national security grounds.

Anthropic said the order covered people outside the United States as well as foreign nationals inside the country, including its own employees who are not U.S. citizens. Because of that reach, the company said it shut off the models for all users.

The company said access to its Claude models was not affected, including Claude Opus 4.8, which Anthropic described as less powerful than the systems covered by the directive. In a post on X, Anthropic apologized to customers, said it believed the government action reflected a misunderstanding and said it was trying to restore access.

Anthropic says the order lacked details

In a company blog post, Anthropic said it received the Commerce Department directive at 5:21 p.m. Eastern time. Anthropic said the letter did not spell out the specific national security issue behind the decision.

Anthropic said officials told the company the order followed the government’s discovery of a way to bypass safeguards in Fable 5. According to Anthropic, those safeguards were meant to keep users from reaching cybersecurity capabilities in Mythos, the underlying model used to build Fable 5.

Anthropic disputed the government’s apparent assessment of the risk. The company said it believed the cited jailbreak was limited to one particular use case, rather than a general method for defeating Fable 5’s protections.

Anthropic also said it believed similar methods could draw comparable cybersecurity capabilities from other publicly available systems, including OpenAI’s GPT-5.5, which it said had not been placed under the same export restrictions. Anthropic argued in its blog post that treating a narrow jailbreak as grounds to recall a commercial model could stop frontier AI providers from launching new systems.

Policy fight widens around Anthropic

Fortune reported that AI policy specialists and industry figures reacted sharply to the directive, which it described as unprecedented. Some interpreted the action as part of a broader dispute between the Trump administration and Anthropic.

Fortune reported that President Donald Trump ordered federal agencies in February to stop using Anthropic models after the company declined the Pentagon’s preferred AI contract terms. Those terms allowed purchased AI systems to be used for any lawful purpose, while Anthropic had sought carve-outs for autonomous weapons and mass domestic surveillance, according to Fortune.

Fortune also reported that the Pentagon labeled Anthropic a supply chain risk in March, forcing the military to stop using its models and barring defense contractors from using them on government work. Anthropic is challenging that designation in federal court, according to Fortune.

Dean Ball, an AI policy expert who briefly served in the Trump administration, criticized the new restriction on X and questioned whether it reflected targeted legal pressure on Anthropic or an aggressive national security stance. Fortune reported that other critics took the opposite view, arguing Anthropic had invited scrutiny by presenting Mythos as unusually dangerous and promoting Fable 5 as a guarded version of it.

Gary Marcus, a longtime AI industry critic, said on social media that the government’s action fit with U.S. efforts to stay ahead of China in advanced AI. According to Fortune, Marcus also said the move could push Chinese-born AI researchers working at U.S. labs to return to China and could make investors question the stability of U.S. AI policy.

This story draws on original reporting from Fortune.