US ban on Anthropic AI model exports strains allied relations

Artificial intelligence has become the latest flashpoint in transatlantic relations after President Donald Trump compelled Anthropic to cut off foreign access to its powerful models, including Mythos 5 and Claude Fable 5, citing national security.

The unprecedented ban, issued last week, applies to all foreign nationals both inside and outside the United States, forcing Anthropic to fully suspend the two models. Prior to the order, Anthropic had granted access to its experimental Claude Mythos Preview to 200 organisations in 15 countries for vulnerability testing. Two public releases were scheduled for early June.

Anthropic said the US government did not explain its reasoning, but understood the Trump administration believed it had found a way to “crack” Fable 5. The ban sent immediate shockwaves through Europe, which relies heavily on US-developed AI.

French President Emmanuel Macron, at this week’s G7 meeting, called the ban a “wake-up call” on AI risks but described the restriction as “the worst thing”. He said: “This reaction is, in a way, nationalist.” Macron stressed the need for cooperation among democracies, warning against “a lack of collaboration among democracies”.

While the US has previously imposed technology sanctions targeting China and Russia, the Anthropic ban also extends to allies with intelligence-sharing and defence treaties with Washington. The decision is a first in the AI sector but follows other transactional policies by the Trump administration, including a global trade war, threats to annex Greenland (an autonomous territory of Denmark), and suggestions of withdrawing from NATO.

Thomas Regnier, spokesperson for the European Commission on technological sovereignty, argued that security is “a shared challenge, not one for a single country”, and that solutions should not “discriminate against partners”. At a closed-door G7 session, countries discussed a “trusted partner” programme for access to cutting-edge AI models, though few details emerged.

The Anthropic ban has amplified calls for greater autonomy among US allies. Canadian Prime Minister Mark Carney said: “No one did anything wrong. But we would be wrong if we simply accepted this without drawing lessons and diversifying.” Former French minister Bruno Retailleau urged Europe to treat AI like nuclear power: “master it or endure it”. Former British minister Tom Tugendhat warned that sovereignty today depends on code more than on cannons.

Marcin Jerzewski, head of the European Values Center for Security Policy’s Taiwan office, said European companies could benefit from the Anthropic case. Before the ban, the German military had already declined to contract with Palantir over data-dependency concerns. German and French intelligence agencies have also turned to European firms instead of Palantir. Attention is now shifting to French AI startup Mistral, the EU’s only domestic rival in frontier models.