When Washington flips the switch on European AI
Anthropic has cut access to its most powerful AI models for non-Americans on Washington's orders. Europe's reaction reveals a dependency no one wanted to name.
At a Glance
Acting on a U.S. government directive citing national security, Anthropic immediately suspended access to its two most advanced AI models — Claude Fable 5 and Mythos 5 — for all non-American users.
Politicians across the French, British, and Dutch political spectrum — from France’s minister for European affairs to far-right Dutch leader Geert Wilders — responded with rare unanimity, calling for technological sovereignty.
The episode crystallizes a reality European strategists had been deferring: the continent’s critical digital infrastructure runs on platforms that foreign governments can shut down overnight.
This image is used for illustrative purposes only.
A switch Washington just tested
On Friday, June 13, 2026, Anthropic — the San Francisco-based AI company co-founded by Dario Amodei, formerly vice president of research at OpenAI — announced it had received a letter from the U.S. government ordering the immediate suspension of access to its Claude Fable 5 and Mythos 5 AI models for all foreign nationals, on grounds of national security.
Anthropic’s own statement, however, went further than the government order as described: the company said it had to disable both models for “our entire customer base” to ensure compliance — phrasing that leaves open whether the shutdown also affects American users, or simply refers to the full population of foreign clients targeted by the directive.
The administration did not specify the precise legal mechanism behind the order. Similar restrictions on advanced AI exports have typically been issued under the Commerce Department’s export-control authority — if that is the case here, it would frame this move less as a new tool and more as an extension of existing U.S. tech-export policy to frontier AI models. This unilateral move could constitute a significant precedent. For the first time at this scale, a government has used its regulatory authority to remotely disable cutting-edge AI tools — not for commercial reasons, but in an explicitly geopolitical logic.
Across party lines, the same diagnosis
What makes this episode remarkable is not only its nature but the breadth of the response it triggered. In France, the United Kingdom, and the Netherlands, figures whose political positions are ordinarily opposed reacted with striking convergence.
Benjamin Haddad, France’s minister delegate for European affairs, called Washington’s decision an “accelerator of the geopolitical battle for AI,” urging Europe to stop being an open market dependent on technologies designed, financed, and controlled elsewhere.
Édouard Philippe — France’s prime minister from 2017 to 2020 and a declared candidate for the 2027 presidential election — offered the most structural diagnosis: AI has now become critical infrastructure, on par with electricity or the internet, and infrastructure that others can unplug.
When a government can remotely disable the AI models running in your hospitals, your research labs, and your businesses, the infrastructure debate is over. The dependency is already here.
On the British side, Al Carns, a Labour member of parliament who resigned as minister for the armed forces earlier in the week over a dispute on the defense budget, broadened the point: this isn’t an AI story — it’s the story of every sector the United Kingdom used to lead. His Conservative colleague Tom Tugendhat was more specific, pointing to the paradox of a country that has prioritized caution over technological opportunity.
In the Netherlands, Geert Wilders, leader of the Partij voor de Vrijheid (Party for Freedom), the country’s hard-right nationalist party, set aside his usual rhetoric to call for accelerated development of autonomous Dutch AI models. The convergence is striking: between the centrist Haddad and the nationalist Wilders, the diagnosis is the same — even if the implied remedies differ.
Jordan Bardella, a member of the European Parliament and president of the Rassemblement National (RN), France’s far-right party, also weighed in, calling Washington’s move a sudden reminder that artificial intelligence is already a major issue of national sovereignty, and urging France to accelerate its support for AI startup Mistral and the broader domestic AI ecosystem.
The mechanics of dependency
What happened Friday illustrates what analysts call “technological leverage risk” — the ability of a state that supplies digital infrastructure to apply pressure on client states by disabling or degrading services those states have come to depend on.
The European Union is not without instruments. The EU AI Act — the artificial intelligence regulation adopted in 2024 and gradually entering into force — is specifically designed to govern AI systems operating on European territory, including those developed abroad. But however ambitious that regulatory framework may be, it does not resolve the underlying question: if the most powerful models are designed, trained, and controlled across the Atlantic, Europe remains fundamentally a consumer, not a producer.
Calls to “Mistral AI” — the French startup founded in 2023 that develops open-weight language models and represents the most visible European push for AI sovereignty — multiplied in Friday’s reactions. Bruno Retailleau, a former French interior minister and 2027 presidential candidate, also cited OVHcloud and Scaleway as building blocks of a sovereign digital infrastructure. These companies exist. They compete effectively in their segments. But they do not yet command the computing power, capital, or scale of the models Washington just switched off.
The bottom line
The question Friday, June 13 actually raises is not whether the Trump administration was right or wrong to flip this switch. It is asking how Europe allowed infrastructure that its own leaders now describe as “critical” — as essential as electricity or the internet, in their own words — to remain entirely in foreign hands for so long.
The rhetorical awakening is real. To be fair, European investment is not nonexistent: Mistral has closed substantial funding rounds, and several governments have floated new compute initiatives. But measured against the scale of spending by U.S. hyperscalers, those efforts remain a fraction of what closing the gap would require. Europe now has a precise date it can point to as the moment it saw the signal. The question is whether, this time, it will decide to act before the next switch is thrown.
Sources: Euronews