Anthropic Says It’s Taking Claude Fable 5 Online to Comply with US Government Law

Anthropic says it is banning two types of AI that launched this week, Claude Fable 5 and Mythos 5to comply with the cargo control instructions received on Friday afternoon from the US government regarding national security.

The unprecedented event is the latest landmark among them Anthropic is the Trump administration. Although the company says the order called for it to stop access to “any foreign national, whether inside or outside the United States, including Anthropic’s foreign employees,” it removed access to all of its customers to ensure compliance.

Earlier this year, Trump’s Department of Defense called Anthropic “Supply chain risk” after the designer Claude wanted to explain clearly how the United States military could use his technology.” The statement prohibited government agencies and contractors from using Anthropic’s technology. to file lawsuits against the Trump administration.

On Tuesday, Anthropic was released publicly Claude Fable 5a version of the company’s Mythos AI model with safeguards that prevent it from answering questions about cybersecurity, biology, and chemistry. Before the public release, which Anthropic said it did in cooperation with the US government, the Mythos Preview AI example it was partially released in April. Its purpose was to provide companies and organizations with the opportunity to use it strong cybersecurity skills to improve their securityI worry that the technology could be used by bad actors to create powerful hacking tools.

In a blog post On Friday, Anthropic said it received a letter from the US government at 5:21pm ET. “The letter did not go into detail about national security,” Anthropic wrote.

“Our understanding is that the government believes it has identified a way to bypass, or ‘jailbreak’ Fable 5,” the company added. “We also reviewed the demonstration of this specific method being used to identify small threats that were previously known. All of these threats appear to be relatively simple, and we found that some common types can also detect them without the need for a bypass method.”

In a blog post, the company said that it has implemented strong security measures to reduce the chances of misuse of Claude Fable 5. Anthropic also said that the jailbreak that the US government found in Claude Fable 5 was minimal, and would not make the attacker any more dangerous than it would have been with another type of AI.

“So far, the government has only given us verbal evidence of a narrow prison, which is not universal, which basically consists of asking the model to read the actual codebase and correct any programming errors,” the company said in its post. “Our understanding is that one potential arrest has been made by the government.”

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

The CEO of Anthropic, Dario Amodei, said in his speech earlier this week that he and the company support a fair, well-organized and transparent government policy that would prevent the release of unsafe AI models. On the company’s blog post Friday, Anthropic said “this does not follow these guidelines.”