The U.S. government blocked European access to its most advanced AI models. Austria's State Secretary for Digitalization, Alexander Pröll, now wants to lure Anthropic to the EU. Meanwhile, China looms as a supposed alternative.

The U.S. recently blocked or delayed access to two of its most advanced AI models for foreign nationals, the latest from OpenAI and Anthropic. For Europe, the takeaway is stark. Access to top-tier AI now hinges on a handful of policymakers in Washington, creating a dependency that could prove dangerous.

"Overnight, the world's largest single market—our EU single market with 450 million people—was cut off from cutting-edge innovation," writes Austrian State Secretary for Digitalization Alexander Pröll (ÖVP) in a letter to EU Commissioner for Technological Sovereignty Henna Virkkunen, seen by the Austrian Kronen Zeitung.

Pröll wants Anthropic's headquarters in Europe

In a LinkedIn post, he lays out his reasoning. "A technology that you don't produce yourself and can only use with permission is not a tool. It is a dependency." Europe should attract Anthropic by offering legal certainty, market access, and capital, all grounded in European values.

Why Anthropic specifically? Pröll points to the company's philosophy. "Anthropic is a company that views the ethical use of AI not as a marketing ploy, but as a core conviction. One that prioritizes safety over speed," Pröll writes in the letter, according to the Kronen Zeitung. "This company would not be constrained in Europe; it would be set free."

The letter only calls for a review, and it's unlikely the Commission will act on it. The initiative isn't realistic. It reads more like an act of desperation from an EU whose dependency built up over 20 years of digital policy failures, not just the last five.

Even if the option existed, Anthropic is unlikely to bite given Trump's retaliatory posture. The company may look cosmopolitan and values-driven by U.S. standards, but it's deeply patriotic. In its Pentagon dispute over AI deployments, Anthropic's main concern was protecting U.S. citizens. Mass surveillance abroad wasn't on the agenda. The models are also reportedly already in use by the NSA.

OpenAI, Anthropic, and others do count on European revenue, though, to fund billion-dollar training runs and data center buildouts. The U.S. market alone might not be enough to cover those costs, and that gives Europe some leverage at the table, even if it's not much.

China looks like a backup, but it's just another leash

AI investor Xiaoyin Qu sketches an alternative scenario. European companies, and for cost reasons, U.S. companies too, could adopt Chinese AI models instead of deepening their dependence on OpenAI and Anthropic. Companies could host Chinese models on their own GPUs, fine-tune them with their own data, and retain control, Qu argues. Trust in Anthropic has already taken a hit after incidents like its handling of Fable, she adds.

Qu also flags a worst case for the U.S. If Chinese open-source models keep gaining market share and get optimized for Huawei chips instead of Nvidia, China could dominate both the model and chip layers. Export controls alone won't fix that. The U.S. would need its own open-source push.

For Europe, though, this scenario just swaps one dependency for another, trading Washington for Beijing. Locally run Chinese AI looks self-sufficient at first glance, but the situation could flip overnight. Open-source models could get new licenses slapped on them, or the best models could be withheld entirely.

The Cold War rhetoric some Chinese AI experts already lean on doesn't suggest generosity ahead. China won't give the EU gifts long-term, only while it hurts the U.S. Real sovereignty requires Europe's own AI and, above all, its own infrastructure.