Anthropic says it will take Claude Fable 5 offline to comply with the US government order

Anthropy says so Two AI models it launched earlier this week, Cloud Fable 5 and Mythos 5, were disabled to comply with an export control directive it received Friday afternoon from the US government due to national security concerns.

This unprecedented incident represents the latest flashpoint between Anthropic and the Trump administration. While the company says the order required it to suspend access to “any foreign national, whether inside or outside the United States, including Anthropian employees who are foreign nationals,” it removed access for all of its customers to ensure compliance.

Earlier this year, Trump’s Department of Defense called Anthropic a “supply chain risk” after Claude’s maker sought to draw red lines on how the US military uses its technology. The designation effectively prevented government agencies and contractors from using Anthropic’s technology. Anthropologie responded by filing lawsuits against the Trump administration.

On Tuesday, Anthropic publicly released Claude Fable 5, a version of the company’s Mythos AI model with safeguards that prevent it from answering questions about cybersecurity, biology, and chemistry. Ahead of the public release, which Anthropic said it conducted in collaboration with the U.S. government, the Mythos Preview AI model had a limited rollout in April. The goal was to give companies and organizations an opportunity to use their strong cybersecurity capabilities to improve their defenses, and eliminate concerns that the technology could be exploited by bad actors to develop powerful hacking tools.

In a blog post on Friday, Anthropic said it received a message from the US government at 5:21 p.m. ET. “The letter did not provide specific details regarding its national security concerns,” Anthropic wrote.

The company added: “Our understanding is that the government believes it has become aware of a method to bypass or ‘jailbreak’ Fable 5.” “We reviewed a demonstration of this specific technique used to identify a small number of previously known minor vulnerabilities. These vulnerabilities all appear to be relatively minor, and we have found that other publicly available models are able to detect them as well without requiring overriding.”

In the blog post, the company argued that it had implemented strong safeguards to reduce the potential for misuse of Claude Fable 5. Anthropic also claimed that the jailbreak the US government found for Claude Fable 5 was narrow, and would not have made the attacker any more dangerous than he would have been with another AI model.

“So far, the government has only given us verbal evidence of a limited, non-universal jailbreak potential, which essentially consists of asking the model to read a specific code base and fix any software flaws,” the company said in its blog. “Our understanding is that one potential jailbreak has been shared with the government.”

Spokesmen for the White House and US Commerce Department did not immediately respond to WIRED’s request for comment.

Anthropic CEO Dario Amodei said in a policy op-ed earlier this week that he and the company support a fair, orderly, and transparent government process that would prevent the release of unsafe AI models. In the company’s blog post on Friday, Anthropic said “this action does not adhere to these principles.”