Europe worries about US AI dominance as tech world gathers in France

The timing felt almost scripted. Just as more than 180,000 attendees prepared to flood into VivaTech in Paris, and as G7 leaders gathered at the lakeside retreat of Evian-les-Bains, the United States quietly tightened access to Anthropic’s most advanced AI models for foreign nationals. Europe arrived at its own showcase having been reminded, once again, that the tools its businesses rely on can be switched off by a decision made in Washington.

Technological sovereignty was always going to dominate both events. Now it dominates with a sharper edge. In the days leading up to these gatherings, policymakers and executives spent their time fretting, in the careful phrasing of the week, about American AI dominance and the glaring absence of credible European alternatives.

This anxiety is not new. What is new is the proof that the dependency is not theoretical. When the US ordered Anthropic to bar foreign nationals from its top-tier systems, the company found the restriction impossible to enforce selectively on a shared cloud infrastructure. So it simply turned the models off for everyone worldwide, including European users who had nothing to do with the directive.

Europe’s answer, such as it is, has one name attached to it more than any other. Mistral has become the company most frequently invoked within the EU’s sovereignty framework, championed by the French government that backs it and criticized by others within the bloc who argue that a single national champion is a fragile foundation for a continent-wide strategy. The Paris-based firm has raised debt to purchase Nvidia chips, committed to building new data centers, and positioned itself as Europe’s primary alternative. Whether one company can bear the weight the rhetoric places on it is the question nobody at VivaTech seems eager to answer out loud.

Beneath the speeches lies a structural problem. Much of Europe’s frontier AI work still runs on American cloud infrastructure, and GPU-as-a-service arrangements often reinforce the very dependency they are meant to alleviate. Renting compute power from a US provider is not the same as owning it, a distinction that becomes stark the moment access policies shift. The EU has planned five AI gigafactory sites and awarded sovereign-cloud contracts, but the build-out is measured in years, while the political risk is measured in news cycles.

At Evian, the conversation took a more transactional turn. Representatives from several G7 countries used the summit’s opening to raise with US Commerce Secretary Howard Lutnick the idea of a “trusted partners” arrangement that would allow allied nations or companies to reach the very American models now being restricted. It is a revealing ask. The European pitch in public is independence; the European request in private, this week, was for a better seat at the American table.

VivaTech itself will look, as it always does, like a continent in confident motion. AWS and Nvidia will showcase French startups in their Startup Village. Robotics demos, voice and decision-intelligence pitches, the usual choreography of a sector that wants to be taken seriously. The choreography is real, and so is the anxiety beneath it. Both can be true at once.

What the week makes clear is that Europe has settled on the diagnosis and is still arguing about the prescription. The dependency is named. The vulnerability is conceded. The gigafactories are planned. The continent now has four days in France to decide whether sovereignty is something you build or something you negotiate. On this week’s evidence, it is hedging on both.