Commerce Secretary Howard Lutnick's June 13 letter to Dario Amodei, published in full by Bloomberg on Monday, is the first time the United States has placed export controls on a deployed commercial AI model. The directive requires government pre-approval before Anthropic can release Fable 5 or Mythos 5 to any foreign national, including foreign nationals inside the country, under threat of criminal and civil penalties. Anthropic has disabled both models worldwide rather than try to comply piecemeal.

The trigger is a single jailbreak.

According to Anthropic's statement, the government has verbally cited exactly one narrow technique: asking Fable 5 to read a specific codebase and fix any software flaws it finds. Anthropic says it reproduced the same vulnerabilities on publicly available models including OpenAI's GPT-5.5, that the flaws surfaced were previously known and minor, and that no tester produced a universal jailbreak across thousands of hours of joint red-teaming with the US government, the UK AISI, and outside parties before the June 9 launch.

The sequencing is its own story. Fortune reports that Amazon CEO Andy Jassy raised concerns with senior administration officials after Amazon researchers used a prompt sequence to extract restricted cyberattack information from a Mythos-class model. White House AI adviser David Sacks posted on X that a trusted partner had identified the jailbreak, that Amodei was asked to fix or withdraw the model, and that he refused. Semafor, cited by Fortune, separately reported government suspicion that a group with ties to China may have obtained access to Mythos. Commerce was on the phone to Anthropic by 1:00 p.m. ET that Friday; the formal Lutnick letter arrived at 5:21 p.m., per Anthropic's own timestamp. The letter, Bloomberg notes, invokes national security authorities but cites no specific incident.

Senior Anthropic technical staff flew to Washington on June 15 for talks with the Trump administration. CNBC's source says the meeting ended without resolution.

Anthropic's written response stakes out the precedent argument directly. "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."

That's the structural piece worth holding onto. Mythos 5 had been gated under Project Glasswing; Fable 5 was generally available to enterprise customers and paid subscribers, with 30-day customer-data retention for forensic monitoring. The directive contains no geographic carve-out, no enterprise exception, and no published threshold for restoration. A competitor's red team, a single posted complaint, and a phone call from Commerce were enough to pull a shipping frontier model offline. Every other lab now knows the procedure.

Sources

https://www.bloomberg.com/news/articles/2026-06-16/read-the-lutnick-letter-that-led-anthropic-to-disable-mythos
https://www.bloomberg.com/news/articles/2026-06-13/anthropic-says-us-limits-foreign-access-to-fable-5-mythos-5
https://www.cnbc.com/2026/06/15/anthropic-mythos-trump-ai.html
https://fortune.com/2026/06/14/how-a-warning-from-amazon-led-the-white-house-to-shut-down-anthropics-mythos-model/
https://www.anthropic.com/news/fable-mythos-access