The Trump administration ordered AI company Anthropic to cut off all foreign access to its latest frontier models last week, a decision that applies the kind of technology restriction typically reserved for adversaries to America's closest allies and accelerates demands for European self-reliance in artificial intelligence.
The order forced Anthropic to take the Mythos 5 and Claude Fable 5 models completely offline. The administration did not provide a written reason, but Anthropic said it understood the government had become aware of a method of "jailbreaking" the Fable 5 model. The company had already shared an earlier preview, Claude Mythos Preview, with 200 institutions across 15 countries for vulnerability testing. The full public release had been planned for early June.
Allies confront an abrupt restriction
French President Emmanuel Macron told a G7 meeting the order was a "wake-up call" but described the limits as a "bad thing" and the response as "strictly nationalist." European Commission tech sovereignty spokesperson Thomas Regnier said addressing AI security is "a shared challenge, not one confined to a single jurisdiction or country," and that solutions should not be "discriminatory against partners." At the same closed-door meeting, G7 leaders discussed a potential "trusted partner" scheme to govern access to advanced AI models.
The Anthropic ban is the first of its kind for AI, but it follows a broader pattern. Over the past 18 months, the Trump administration has launched a global trade war, threatened to withdraw from NATO, and linked weapons supplies to Ukraine with European help in reopening the Strait of Hormuz. An earlier Biden-era "small yard, high fence" chip export scheme was scrapped by Trump in May 2025, and the administration later approved Nvidia H200 chip sales to a limited group of Chinese firms.
Calls to treat AI as a sovereignty question
Canadian Prime Minister Mark Carney said ahead of the G7 summit that over-reliance on US technology leaves allies vulnerable. "Nobody has done anything wrong in the situation. But we will have done something wrong if we just accept this, don't take the lesson, don't build out and diversify," he said. French presidential candidate Bruno Retailleau compared AI to nuclear power: "We must treat AI the way we treated nuclear power: we must think of it as part of our sovereignty. Master it or suffer it: there is no other path." Former UK security minister Tom Tugendhat said sovereignty is increasingly "more about code than cannons."
European tech firms benefit from push to reduce dependence
Even before the ban, German and French spy agencies had shifted contracts away from US data analytics firm Palantir toward European alternatives to limit access to national data systems. Paris-based Mistral remains the EU's only major homegrown frontier-model competitor, and analysts say the Anthropic incident could accelerate demand for its services. "European governments are growing uneasy about their overreliance on US-controlled technologies," said Marcin Jerzewski of the European Values Center for Security Policy. "To this end, European companies might benefit from the Anthropic incident."
Why this matters for government professionals
The Anthropic ban demonstrates that access to advanced AI can shift overnight based on a single administration's security calculus. For civil servants, defence planners and policy makers, it shows the urgency of understanding AI supply chains and developing domestic alternatives. Building the expertise to evaluate these risks is now a core governance skill. Frameworks for AI for Government offer structured ways to think about sovereignty and service continuity, while an AI Learning Path for Policy Makers can provide the technical and strategic grounding required to make informed decisions in an environment where code is treated as a strategic asset.
Your membership also unlocks:
