On June 12, Anthropic announced that it was disabling access to two of its advanced artificial intelligence large language models, Claude Fable 5 and Mythos 5, after receiving an order from the U.S. government to suspend access to these models by any non-U.S. national.
Anthropic described the order (which the U.S. government has not publicly disclosed) as an “export controls directive” that cites “national security authorities” as the basis for the suspension. Anthropic announced that the U.S. government’s concerns relate to reported methods for bypassing, or “jailbreaking,” the safety controls deployed by Anthropic to prevent misuse of its models for illicit activities. According to Anthropic, the particular vulnerabilities at issue are minor and are common to other AI models.
The U.S. government has not publicized the order that it issued to Anthropic, the reason that the order was issued, or the legal basis for the order. Based on media reports that the directive was reportedly issued by U.S. Secretary of Commerce Howard Lutnick, the legal authority for the order is most likely the Export Controls Reform Act of 2018 (“ECRA”), which is the Commerce Department’s primary authority for administering U.S. “dual-use” export controls.
Among other things, ECRA authorizes Commerce to privately “inform” U.S. companies like Anthropic that an export license from Commerce is required to engage in any activity that Commerce finds may involve the development of weapons of mass destruction. Commerce has used this authority fairly regularly with respect to exports of semiconductors and semiconductor equipment to China by issuing so-called “is-informed” letters to U.S. exporters imposing export license requirements with specific counterparties that may be involved in WMD-related activities. These letters, which the U.S. government does not make public, are often precursors to broader, public export restrictions on specific parties in China or elsewhere.
While Commerce’s use of its “is informed” authority is not novel, the breadth of the order issued with respect to the Anthropic models is unprecedented. Access to these models, which previously was not subject to any export controls restrictions, is now apparently off limits for any foreign national, anywhere in the world – including persons located in the United States.
This order is a remarkable contrast with the Trump Administration’s heretofore loud and unapologetic “hands off” approach to export restrictions on U.S. AI models. In May 2025, Commerce announced that it would rescind the Biden Administration’s “AI diffusion” export controls rule because it “would have stifled American innovation and saddled companies with burdensome new regulatory requirements.” An executive order issued earlier this month touted success in achieving “tremendous technological growth and economic investment in AI” through “slashing the bureaucratic constraints that the prior administration placed on America’s AI developers and researchers.”
It is also not clear why Anthropic’s models were singled out for this treatment, as compared to other U.S. large language model AI companies. This is leading some to speculate – in light of Anthropic’s designation as a supply chain risk by the Department of War in February — that Anthropic is being specifically targeted for scrutiny by the U.S. government.
Commerce can end speculation about its motivations and concerns with respect to Anthropic by issuing public guidance on the nature of the threat it is addressing through the Anthropic order, the steps that must be taken to alleviate the threat, and the likelihood that other AI models may face similar vulnerabilities (and presumably similar restrictions). Unless the U.S government is willing to consider a broader set of safety-related regulations around AI technologies, however, Commerce may have no options other than to deploy one-off, private directives like the one issued to Anthropic – which leaves industry and the American public in the dark as to the true nature of the threat being addressed.
Editor’s note: Readers may also be interested in Justin Hendrix’s The Mythos Recall and Washington’s Missing AI Safety Playbook, Just Security, June 13, 2026
