The US Commerce Department ordered Anthropic to obtain US approval for foreign nationals to use its Claude Fable 5 and Mythos 5 AI models, effectively banning foreign access worldwide starting June 12, 2026 [1]. The department cited national security risks linked to "jailbreaking" techniques that bypass AI safety measures as justification for the export controls [1, 2].

Anthropic suspended global access to the models following the government restrictions [1, 2]. The AI startup had released Claude Fable 5 and Mythos 5 just days before the controls were imposed in early June 2026 [1, 2]. Prior to the ban, Anthropic had granted 200 institutions across 15 countries access to the Claude Mythos Preview model for vulnerability testing [2].

The US export ban marks an unprecedented use of export control laws, as it targets the use—not just the transfer—of US-origin AI technology [3]. The restrictions extend beyond traditional targets like China and Russia, also affecting allies of the US [1, 2].

French President Emmanuel Macron criticized the US decision at the G7 summit on June 19, calling the reaction "strictly nationalist" and saying the limits were a "bad thing" that could undermine trust in US AI offerings [1, 2]. Risto Uuk, head of European policy at the Future of Life Institute, called the US move "hasty and uninformed," urging Washington to implement clearer AI regulations akin to the EU’s upcoming framework [1].

The Trump administration has not publicly released detailed evidence supporting the severity of jailbreaking risks. Anthropic downplayed the risks, calling them minor and overblown and noting similar vulnerabilities affect rival AI platforms [1, 2].

Anthropic, which filed plans to go public later in 2026 and potentially raise billions, now faces restrictions that may impact its global business prospects [1]. The company’s next step will be working through US approval processes for foreign users and assessing how the export controls affect its launch plans.