The American artificial intelligence start-up Anthropic announced on Friday that it was suspending access to its two most powerful models, Fable 5 and Mythos 5. It is thus complying with a directive from the American government citing “national security”, just three days after their commercial launch.
Washington ordered, under export control, to cut off access to these models for “any foreign national, inside or outside the United States”, including “foreign employees” of Anthropic, according to the company press release. Unable to sort its users according to their nationality, Anthropic, already in dispute with the Trump administration, announced that it would have to “brutally deactivate” the two models for all of its customers in order to comply.
The company reports receiving a government directive at 5:21 p.m., “which did not specifically detail its national security concerns.” According to the American media Axios, the directive comes from Commerce Secretary Howard Lutnick. He apparently made this decision after learning that a company using the models had managed to circumvent the safeguards theoretically preventing malicious use.
“A misunderstanding”
Asked by AFP, the Ministry of Commerce did not immediately respond. “We dispute that the discovery of a potential circumvention” of the security measures around Fable 5 “justifies the recall of a commercial model deployed to hundreds of millions of people”, writes Anthropic, which wants to see in the affair “a misunderstanding”.
“If this standard were applied to the entire sector, we believe it would essentially halt all new deployments of cutting-edge AI models,” adds the company, which says it is working to restore access “as soon as possible”. Launched on Tuesday, Fable 5 is the first model in the Mythos class, Anthropic’s most advanced range that the company unveiled in April without opening it to the public for security reasons.
Unprecedented speed and acuity
Fable 5 is restricted in sensitive areas such as cybersecurity and the risks of biological and chemical attack. Its unrestricted version, Mythos 5, reserved for some 200 companies, organizations and state agencies, is presented as capable of detecting and exploiting security vulnerabilities at unprecedented speed and acuity.
The issue of security, which Anthropic has made one of its main commercial arguments, has already dragged the San Francisco start-up into a standoff with the Trump administration, which led the Pentagon to terminate its contracts with the company. This suspension comes as Anthropic, like its rival OpenAI, announced in June that it had filed its IPO application.
