- Commerce Secretary Lutnick sent Anthropic CEO Dario Amodei a letter last Friday ordering the company not to provide Fable 5 or Mythos 5 to any foreign national worldwide without an individually-validated export license from the Commerce Department.
- The letter threatened criminal and civil penalties, cited US export control statutes covering civilian tech that could serve an adversary’s military intelligence purposes, and gave no public rationale.
- Anthropic disabled both models immediately; technical staff then held virtual and in-person meetings at Commerce with US officials about specific security vulnerabilities in Fable 5’s guardrails.
- Anthropic called the action disproportionate: “If this standard was applied across the industry, we believe it would essentially halt all new model deployments for all frontier model providers.”
What Happened?
Bloomberg obtained the letter Commerce Secretary Howard Lutnick sent to Anthropic CEO Dario Amodei last Friday, revealing the legal mechanism used to force the company offline. The letter ordered Anthropic to treat Fable 5 and Mythos 5 as export-controlled technology, requiring a license before granting access to any foreign national anywhere in the world. Lutnick cited US statutes permitting export controls on civilian technology that could be used for intelligence purposes by an adversary’s military, and threatened both criminal and civil penalties for violations. Anthropic disabled all access late Friday. The company has since held virtual and in-person meetings at Commerce. The underlying trigger: Amazon researchers found a way to bypass Fable 5’s safety guardrails shortly after launch and alerted administration officials.
Why It Matters?
This is the most significant direct US government intervention into an AI company’s commercial operations on record — applying export controls to a consumer product deployed to hundreds of millions of users. Anthropic publicly disagreed, arguing a “narrow potential jailbreak” shouldn’t warrant a global recall, and warned the standard would freeze all frontier AI deployments industry-wide. The episode arrives as Anthropic filed confidentially for an IPO at a valuation topping $900 billion, and every day Fable 5 and Mythos 5 remain offline compounds both financial and competitive damage. The episode also puts Amazon in an awkward position: it was Amazon researchers who surfaced the vulnerability to the government, even as Amazon is Anthropic’s largest investor and cloud compute partner.
What’s Next?
Anthropic must formally apply for an individually-validated export license — typically a weeks-to-months process — unless a negotiated resolution comes first. Both sides have signaled urgency. The outcome will set a precedent for how far Washington is prepared to go in imposing national security restrictions on frontier AI model deployments. For Anthropic’s IPO, investors will need to assess whether export control risk is now structural rather than a one-time event.
Source: Bloomberg
