At 5:21 p.m. ET on June 12, a Commerce Department letter landed at Anthropic ordering the company to suspend global access to Claude Fable 5 and Claude Mythos 5, and by the end of the night both models were dark. Per Anthropic's own statement, it's the first export-control action ever taken against a deployed US frontier model. Fable 5 had been live for three days.

The trigger, according to a verbal description the government provided without specific technical detail, is a jailbreak that "essentially consists of asking the model to read a specific codebase and fix any software flaws." Anthropic calls it narrow and non-universal, and says its review of what it believes is the underlying report shows the same capability is widely available from other frontier models, including OpenAI's GPT-5.5.

That's the analytical center of the story. A model serving hundreds of millions of users has been pulled off the market over a capability its maker says ships unguarded in a direct competitor.

Anthropic's posture is unusually pointed for a company under a federal directive: "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." The company notes that Fable, the guard-railed deployment of the cybersecurity-capable Mythos 5, went through thousands of hours of red-teaming with the US government and the UK AI Safety Institute, that no tester produced a universal jailbreak, and that Fable's architecture relies on independent classifiers operating outside the model, so a refusal bypass doesn't automatically open the hardened cyber and bio categories. A 30-day customer-data retention policy exists precisely to support post-hoc forensics, and Anthropic says it has received no disclosure of a non-universal jailbreak producing a harmful real-world result. Vals AI had ranked Fable 5 the most capable publicly accessible model at release.

The political subtext is louder than the technical one. Fortune reports that Amazon's CEO escalated the Fable 5 concern directly to the White House. Anthropic was already in federal court contesting a March Pentagon designation as a "supply chain risk," a label that bars defense contractors from using Claude after the company refused contract language permitting use "for any lawful purpose" and sought carve-outs for autonomous weapons and mass domestic surveillance. In February, President Trump had ordered federal agencies off Anthropic models. Earlier this month the company confidentially filed for an IPO at a $965 billion valuation. Former AI czar David Sacks has accused it of "regulatory capture" and "fear-mongering."

Cybersecurity researcher Peter Girnus offered the cleanest read of the dynamic: "If you describe your product as a munition in every press release, eventually a government takes you at your word."

A safety-pilled lab built its brand on the premise that its models were uniquely dangerous. The state has now treated that brand as a regulatory finding.

Sources

https://www.anthropic.com/news/fable-mythos-access
https://techcrunch.com/2026/06/12/anthropics-safety-warnings-may-have-just-backfired-the-government-has-pulled-the-plug-on-its-most-powerful-ai/
https://fortune.com/2026/06/14/how-a-warning-from-amazon-led-the-white-house-to-shut-down-anthropics-mythos-model/
https://www.bloomberg.com/news/articles/2026-06-13/anthropic-says-us-limits-foreign-access-to-fable-5-mythos-5
https://www.cnbc.com/2026/06/12/anthropic-disables-access-to-fable-5-and-mythos-5-to-comply-with-government-directive.html