Commerce Department forces Anthropic to disable two AI models for all foreign nationals in a first-of-its-kind export-control order

On June 12, 2026, the Department of Commerce issued an export-control directive ordering Anthropic to suspend all access to two of its most advanced AI models, Fable 5 and Mythos 5, for any foreign national — anywhere outside the United States and any foreign person within it, including Anthropic's own foreign-national employees. The order came as a letter from Commerce Secretary Howard Lutnick to Anthropic's chief executive, Dario Amodei, and cited national-security authorities. Because the company could not cleanly segregate access, complying meant disabling both models for all customers worldwide; Anthropic's other models were unaffected.

The government's stated concern, as relayed by Anthropic, was that a method of "jailbreaking" Fable 5 had been identified, though the letter provided no specifics. Anthropic said it reviewed the cited technique and found that it surfaced only minor, previously known vulnerabilities that other publicly available models — it named OpenAI's GPT-5.5 — could also discover, and that it had disclosed at launch that no model is fully resistant to narrow jailbreaks. The company said it was complying with the legal directive while disagreeing that a narrow, disputed vulnerability justified recalling a model deployed to hundreds of millions of people, and warned that applying that standard across the industry would halt frontier deployments.

Contemporaneous reporting described the action as unprecedented: the United States had used export controls before to restrict the semiconductor chips that power AI, but not the models themselves, and no leading American AI company had previously been forced to take a publicly deployed model offline at the government's direction. The directive also did not arrive in a vacuum. The administration had spent months in an open conflict with Anthropic that began when the company refused, in contract talks, to give the Pentagon unrestricted access to its models; President Trump ordered federal agencies and contractors to stop doing business with the company and called its stance a "disastrous mistake," and the Pentagon designated Anthropic a "supply chain risk" in March.

Export controls are a legitimate national-security tool, and a government may act on genuine concerns about a model's capabilities. What places this action in the archive is the convergence of factors around it: a first-of-its-kind, single-company order, issued on a basis the government did not disclose and that the company says is minor and not unique to its models, against a firm the administration had already moved repeatedly to punish. Recorded as part of the campaign documented in this episode, the entry treats the abuse as the targeted, opaque use of regulatory power against a disfavored company — not the existence of export authority itself. Whether the security rationale is genuine remains disputed, and further independent reporting may sharpen the picture.