US orders Anthropic to block foreign access to its most advanced AI models, company switches them off for everyone

The order and the blackout

On Friday, 12 June, US Commerce Secretary Howard Lutnick directed Anthropic to suspend access to its most advanced models, Mythos 5 and Fable 5, for any foreign national. The official rationale was national security, though details were sparse. Instead of implementing a selective block, the San Francisco startup announced it would "immediately deactivate" the two models for every user worldwide, effectively pulling its frontier AI offline.

Europe cries foul

European Commission spokesperson Thomas Regnier warned that emergency restrictions on cutting‑edge AI must be applied in a non‑discriminatory way towards allies. He stressed that the episode "further underlines Europe's need for technological sovereignty" and pointed to existing bloc‑level tools (the AI Act, the Cyber Resilience Act and the NIS2 directive) that are designed to manage precisely this kind of risk. Brussels is examining the concrete consequences for European users of Anthropic’s services.

“We think that emergency measures adopted in this sense should not be discriminatory towards partners.”

Sovereignty push and the French presidential race

The decision injected a sharp sovereignty angle into French politics, where the next presidential election is approaching. Far‑right leader Jordan Bardella, the current poll front‑runner, declared that the episode "reminds us that artificial intelligence is already a major issue of national sovereignty" and called for accelerating support for Mistral AI, the French champion. Far‑left figure Jean‑Luc Mélenchon struck a similar chord: "Trump's decision concerning Anthropic proves the urgency of being independent and sovereign."

“Nations that do not quickly develop their own models will always depend more and more on the choices of other powers.”

In the European Parliament, Finnish conservative MEP Aura Salla said Europe "cannot keep building its tech stack on access that can be switched off overnight by a foreign government," while Bulgarian lawmaker Eva Maydell insisted the bloc must identify which capabilities to build urgently. Much of the reaction spotlighted Mistral, which is reportedly in talks to raise €3 billion at a €20 billion valuation.

Tumultuous history

Anthropic’s relationship with the Trump administration had already been rocky. In February 2026, the White House ordered all federal agencies to stop using Anthropic’s technology immediately; that directive was later suspended by a federal judge. The company’s CEO, Dario Amodei, has now challenged the current export‑control order in court. Unrelatedly, the outlines of the Mythos model were exposed in late March after a human error placed internal data on a publicly accessible server. The model was already noted for its strong cybersecurity capabilities, making the leak and subsequent blackout especially sensitive.

“Europe cannot keep building its tech stack on access that can be switched off overnight by a foreign government.”