At 5:21 p.m. Eastern on Friday, a Commerce Department letter landed at Anthropic ordering the company to suspend access to its two most capable models, Fable 5 and Mythos 5, for any foreign national. Anthropic complied within hours, then issued a written dissent calling the action a misuse of authority.

Fable 5 had been live for three days. It was Anthropic’s first broadly available frontier-tier model, shipped with new safeguards designed to block responses in specific high-risk areas, and according to the company’s own statement, deployed to hundreds of millions of people. Mythos 5 had never been broadly available at all; it was kept inside Project Glasswing, a tightly restricted program covering roughly 50 vetted organizations including Amazon, Apple, Google, Microsoft, and CrowdStrike.

The directive, per Anthropic, offered no specific national-security detail. The company’s read: officials had learned of a jailbreak of Fable 5, reviewed a demonstration, and identified a small number of previously known, minor vulnerabilities. Anthropic noted, pointedly, that the same jailbreak would elicit comparable capabilities from OpenAI’s GPT-5.5, which isn’t subject to similar export controls.

“the finding of a narrow potential jailbreak,” the company wrote, shouldn’t justify recalling a model deployed to hundreds of millions of people. The precedent, it argued, “would essentially halt all new model deployments for all frontier model providers.”

The political backdrop is the real story. In February, the administration ordered federal agencies to stop using Anthropic models after the company refused Pentagon contract language allowing its models to be used “for any lawful purpose,” including autonomous weapons and mass domestic surveillance. In March, the Defense Department labeled Anthropic a supply chain risk after contract negotiations collapsed, and Anthropic sued the Trump administration to reverse the designation. Earlier this month, the company confidentially filed for a public listing at a $965 billion valuation.

There’s also the matter of the company’s own marketing. In April, the Claude Mythos Preview captivated Wall Street and government officials with cybersecurity demonstrations that Anthropic framed as proof of the model’s danger. That same month, Sam Altman told podcaster Ashlee Vance the posture amounted to “fear-based marketing.” The Commerce Department appears to have taken the marketing at face value, and decided to act on it.

Sources