U.S. Government Export Controls Force Anthropic to Pull Frontier AI Models, Exposing New Realities of State Oversight

The U.S. Commerce Department, under Secretary Howard Lutnick, issued a national security export control directive on June 12, 2026, ordering Anthropic to suspend access to its advanced models Fable 5 and Mythos 5 for any foreign national—inside or outside the U.S., including the company's own employees. Anthropic's official statement confirms the order's broad reach forced the company to disable the models for all users, as selective filtering proved impossible in real time.[1]

This marks the first documented instance of federal intervention compelling a leading AI firm to withdraw a publicly deployed frontier model. Axios and CNN reported the directive stemmed from concerns over cybersecurity vulnerabilities, specifically methods to jailbreak Fable's safeguards, though the government provided no detailed rationale.[2][3] Politico detailed a frantic 24-hour White House effort to persuade Anthropic to voluntarily withdraw the models before imposing controls.[4]

Impacts extended beyond commercial users: Korean partners including the Korea Internet & Security Agency and Samsung, plus European entities like NATO and ENISA via Glasswing programs, lost access. The NSA's separate access to Mythos for offensive cyber operations under Project Glasswing appears unaffected. Anthropic has dispatched technical staff to Washington for resolution talks, with sources indicating both sides seek a swift fix.[5]

The episode illustrates concrete corporate disruption from treating advanced AI as a national security asset, aligning with broader U.S. export control trends previously applied to semiconductors and now extending to deployed models.