Europe must forge its own AI sovereignty path against American control

The decision by the Trump administration to force Anthropic, one of the world's leading artificial intelligence (AI) companies, to suspend access to its most advanced models, Mythos 5, and its consumer version, Fable 5, marks a dramatic turning point in AI governance. Unprecedented in both form and suddenness, the move reveals a unilateral interventionism that creates lasting uncertainty about the deployment of the most powerful technologies.

Officially, Washington cited risks to national security. Anthropic's models can detect the most difficult cybersecurity vulnerabilities at unprecedented speed. If used by foreign actors, whether hostile or not, these models are now considered incompatible with American strategic interests.

This argument is not new, but the method raises questions. By forcing a private company to freeze its language models, without a clear framework or prior consultation with the broader ecosystem, the Trump administration has turned a regulatory issue into a show of political force. The US government is sending an unambiguous message: it reserves the right to halt innovation at any time, whenever it deems necessary.

This is a warning for Europe; a stark reminder that digital dependence is not only economic but also political. When key technologies are designed, hosted and controlled elsewhere, access to them can be challenged by a foreign decision based on national priorities that are not our own. The forced suspension of American models highlights what could become a future rupture in algorithmic supply.

The European Union has implemented a demanding framework with the AI Act, which is sometimes criticized for being too cumbersome. But what happened with Anthropic shows the opposite extreme: regulation by fits and starts, unpredictable and dependent on the political climate. Between these two models, Europe must chart its own course: investing in its capabilities, supporting and coordinating its champions, and taking control of its infrastructure. The first step was taken with the European Commission's legislative package presented on June 3. But progress remains too timid. The direction is the right one, but the pace of execution is too slow, and the scale of the measures still too limited.

French political leaders appear stunned by the situation. With rare unity, parties across the political spectrum condemned a dependence that has become untenable and called for a "European awakening," without specifying what form this should take or what resources should be committed. The prevailing sense is that the debate is only just beginning, while time is running out.

It is urgent to move beyond this indignant unanimity. The 2027 presidential election campaign is an opportunity to move past mere rhetoric. Candidates should concretely state their plan: what resources they would allocate for sovereign computing, what investments in competitive European models, what strategy to diversify our critical infrastructure,and what doctrine to adopt in the face of American power. The Anthropic episode should not remain merely a technological incident. It must become the decisive argument for political resolve to finally rise to meet the sovereignty challenges facing both France and Europe.