At 5:21pm ET on June 12, Anthropic received a Commerce Department export-control directive ordering it to block all foreign-national access to Claude Fable 5 and Claude Mythos 5, including its own non-citizen employees. With no mechanism to filter customers by nationality in real time, the company shut both models off for everyone, worldwide, within hours. Fable 5 had been public for three days.

This is the first time the federal government has invoked export controls against a frontier model from a US lab, and it lands on a company whose entire policy posture has been built around volunteering that its models are dangerous. Anthropic ran thousands of hours of pre-launch red-teaming with the US government, UK AISI, and third-party organizations. Mythos 5, the unrestricted sibling, was already gated behind Project Glasswing, a closed-access program with roughly 50 vetted organizations including Amazon, Apple, Google, Microsoft, and CrowdStrike. Fable 5 is, in Anthropic's own framing, the "safeguarded sibling" of the same underlying model, wrapped in classifier-based blocks for cybersecurity and biology prompts. It launched with the top score on the Vals AI benchmark suite among publicly available models.

The directive didn't name the national-security concern. Officials told Anthropic the action followed disclosure of a jailbreak. Anthropic reviewed what it believes is the underlying report and described the technique as narrow: prompting the model to read a specific codebase and identify software flaws, a capability the company says is "widely available from other models (including OpenAI's GPT-5.5)" and used routinely by defensive security teams. Red-teaming, Anthropic says, produced no universal jailbreak, and the company hasn't received any disclosure of a non-universal jailbreak that produced a harmful result.

The corporate statement is unusually pointed. "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. If this standard was applied across the industry, we believe it would essentially halt all new model deployments for all frontier model providers."

The context is what makes the action legible. In early March, the Pentagon declared Anthropic a "supply chain risk," a label historically reserved for foreign adversaries, barring defense contractors from using Claude on government work. Anthropic is contesting that designation in federal court. Earlier this month, the company confidentially filed for an IPO at its most recent valuation of $965 billion. Access to Claude Opus 4.8 and other models is untouched.

Dean Ball, an AI policy researcher and former Trump-administration appointee, posted on X that he couldn't tell "if this is lawfare against Anthropic in particular or extreme national-security hawkery." Both readings assume the safety case is pretextual. That's the structural problem with spending years telling regulators your models are uniquely dangerous: eventually one of them takes you at your word, on a timeline of their choosing.

US Commerce Department forces Anthropic to pull Fable 5 and Mythos 5 worldwide, three days after launch

Sources