The US government forced Anthropic to pull its two most advanced AI models offline Friday evening, marking the first time a leading AI company has taken a publicly deployed model down due to federal intervention.
Commerce Secretary Howard Lutnick issued an export control directive at 5:21 p.m. ET ordering Anthropic to suspend access to Fable 5 and Mythos 5 for any foreign national, including non-US citizens employed by the company. To guarantee compliance, Anthropic disabled both models for all customers worldwide. The directive arrived just days after Anthropic launched Fable 5 and Mythos 5 on Tuesday, touting them as state-of-the-art across multiple industry benchmarks.
Fable 5, a "Mythos-class model" with safety guardrails for general use, was the company's first release of such an advanced system to the public. The government cited a jailbreak technique that could trick Fable 5 into scanning a specific codebase to identify and patch software vulnerabilities. But Anthropic disputes the basis for the action.
"To date, the government has only given us verbal evidence of a potential narrow, non-universal jailbreak," the company said in a statement. Anthropic reviewed a demonstration of the technique and found it identified "a small number of previously known, minor vulnerabilities" that other publicly available AI models can also discover without any bypass.
"We disagree that the finding of a narrow potential jailbreak should be cause for recalling a commercial model deployed to hundreds of millions of people," Anthropic added. The company warned that applying this standard across the industry "would essentially halt all new model deployments for all frontier model providers." The order stems from Mythos, Anthropic's flagship model that has rattled both Wall Street and the US government with its cybersecurity capabilities.
In April, Anthropic limited Mythos Preview to a select group of partners through Project Glasswing, a cybersecurity initiative. By last week, the Financial Times reported that the National Security Agency was using Mythos to conduct offensive cyberattacks.
Friday's directive is the latest chapter in a contentious relationship between Anthropic and the Trump administration. In February, the Pentagon declared Anthropic a "supply chain risk" after the company pushed for safety guardrails on military use of its AI.
President Donald Trump posted on Truth Social that "The Leftwing nut jobs at Anthropic have made a DISASTROUS MISTAKE trying to STRONG-ARM the Department of War." Anthropic sued and a federal judge in California ruled in the company's favor, though the case remains active.
Despite the tensions, Anthropic had worked with the administration on the recent executive order on AI, signed June 2, which asks companies to voluntarily share advanced models with the government before broader release. Sources familiar with the situation told CNN that Anthropic executives had been invited to the White House for a signing ceremony that was canceled at the last minute.
Axios reported the directive requires Anthropic to obtain a license for "the export, re-export or domestic transfer" of the models. The company said it hopes to reinstate access as soon as possible, calling the order the result of "a misunderstanding."
