.png)
The US Commerce Department on June 13, 2026, gave Anthropic a 90-minute ultimatum: restrict Claude Fable 5 and Mythos 5 to US citizens only, or face further consequences. The directive, signed by Secretary Howard Lutnick, invoked national security authorities. Rather than implement nationality-based access controls on the fly, Anthropic chose to pull both models entirely from global deployment.
The order stemmed from two allegations: that Anthropic had provided access to Mythos 5 to a suspected China-linked company, and that the model contained what officials described as a "potential, narrow, non-universal jailbreak." Anthropic disputed the government's reasoning in a public statement, arguing that the same alleged capability was already accessible through other publicly available models, including OpenAI's GPT-5.5, and characterizing the jailbreak concern as overstated.
According to the Washington Post, Trump administration officials had been weighing export controls on Anthropic's models for weeks before issuing the ultimatum—well before the public releases. The incident marks the first time the US government has used an export control directive to force a company to take a live, publicly deployed AI model offline. House Science Committee Ranking Member Zoe Lofgren condemned the move as an "attack on the AI sector."
Senior Anthropic leaders traveled to Washington, DC, on June 15 to meet with White House officials in an effort to resolve the standoff. The Information reported that no agreement was reached in those talks, though both sides indicated a desire to reach a resolution quickly. As of June 16, Anthropic's two most powerful models remain offline for all users, and no timeline for reinstatement has been announced.