Anthropic Announced It Will Take Claude Fable 5 Offline to Adhere to US Government Mandate

Anthropic has announced that it is shutting down two AI models introduced earlier this week, Claude Fable 5 and Mythos 5, in response to an export control order issued Friday afternoon by the US government, citing national security issues.

This unprecedented event represents the latest tension between Anthropic and the Trump administration. The company claims the directive required it to restrict access for “any foreign national, whether inside or outside the United States, including foreign national Anthropic employees.” Consequently, it has revoked access for all customers to ensure compliance.

Earlier this year, the Department of Defense under Trump classified Anthropic as a “supply chain risk” after the Claude developer attempted to establish limits on how its technology could be utilized by the US military. This designation effectively prohibited government agencies and contractors from employing Anthropic’s technology. In response, Anthropic filed lawsuits against the Trump administration.

On Tuesday, Anthropic launched Claude Fable 5, an iteration of the Mythos AI model, featuring safeguards to prevent it from addressing inquiries related to cybersecurity, biology, and chemistry. Prior to this public launch, which Anthropic stated was conducted in collaboration with the US government, the Mythos Preview AI model had undergone a limited rollout in April, aimed at enabling businesses and organizations to enhance their cybersecurity capabilities and address concerns about potential misuse of the technology by malicious actors.

In a blog post on Friday, Anthropic noted it received a letter from the US government at 5:21 PM ET. “The letter did not provide specific details of its national security concern,” the company remarked.

“We understand that the government believes it has become aware of a method to bypass, or ‘jailbreak’ Fable 5,” the company stated. “We reviewed a demonstration of this technique being used to identify a few known, minor vulnerabilities. These vulnerabilities are relatively straightforward, and we found that other publicly available models can also discover them without requiring a bypass.”

In its blog post, the company contended that it has instituted robust safeguards to minimize the potential misuse of Claude Fable 5. Anthropic also claimed that the jailbreak identified by the US government was limited in scope and would not render an attacker significantly more dangerous than they would be with another AI model.

“To date, the government has only provided us with verbal evidence of a potential narrow, non-universal jailbreak, which essentially involves asking the model to read a specific codebase and rectify any software flaws,” the company indicated in its blog post. “Our understanding is that one potential jailbreak was communicated to the government.”

Spokespersons for the White House and the US Commerce Department did not promptly respond to WIRED’s request for comments.

Anthropic CEO Dario Amodei expressed in a policy essay earlier this week that both he and the company advocate for a fair, structured, and transparent government process that would prevent the release of unsafe AI models. In the blog post released on Friday, Anthropic asserted that “this action does not adhere to those principles.”