US ban on Anthropic's top AI models for foreigners jolts Europe into sovereignty push

Anthropic, a leading American AI firm already blacklisted by the Pentagon, was compelled by the Trump administration to disable access to Claude Mythos 5 and Claude Fable 5 for non-U.S. nationals on Friday, 12 June 2026. The move was justified on national security grounds and represents the second such restriction in under two months. The sudden cutoff sent tremors through European governments and tech circles, reviving a long-smoldering debate about the continent's digital dependency on the United States.

A brutal wake-up call

The decision caught European users and enterprises off guard. Anthropic, which is preparing an initial public offering alongside rival OpenAI, saw its valuation triple in four months before the ban. The forced gating of its most capable language models was immediately framed in Paris as another reminder of the vulnerabilities created by reliance on American technology, echoing earlier disruptions such as the Starlink dependency episode in Ukraine. French presidential candidates and incumbent officials alike called for a rapid awakening.

The Trump administration's decision makes us realize the extent of European dependence.

France answers with cash and a rallying cry

On the eve of the VivaTech conference in Paris, the French government announced an additional €655 million to accelerate the deployment of artificial intelligence in public administration, healthcare, justice and defense. The funding package was explicitly linked to the Anthropic shock. The episode is expected to dominate the 10th edition of VivaTech, where digital sovereignty will be one of the central themes. European champions such as Mistral AI are now cast as strategic assets rather than mere commercial ventures.

The scale of the dependency

The raw numbers explain the alarm. Multiple studies cited in French media show that the United States controls roughly three-quarters of global computing capacity, while Europe accounts for about five percent. The most advanced AI models, the cloud infrastructure that runs them, and the cutting-edge chips they require remain overwhelmingly American.

A global investment race

American AI giants are projected to invest more than $700 billion in 2026, compared with about $100 billion by Chinese firms, according to a report in Les Echos. Chinese models, however, can be up to twenty times cheaper. This cost asymmetry complicates the picture for American shareholders and European policymakers alike.

G7 lunch overshadowed

The affair has injected itself into the G7 summit agenda. A working lunch between G7 leaders and the heads of major AI companies is scheduled for Wednesday in Evian. The meeting, originally meant to discuss governance and cooperation, now carries a sharper edge following what POLITICO describes as a "shockwave" that has rattled capitals far beyond Washington.