Anthropic restriction fuels Europe 2031 AI sovereignty debate

When the US government ordered Anthropic to restrict access to its most advanced AI models for non-US users, a fictional scenario suddenly looked less fictional. The Europe 2031 report, published in June, had projected exactly that event for 2029. It arrived within days. A Tech Policy Press podcast hosted by Ramsha Jahangir uses that gap to stage the argument European policy circles are actually having right now.

On one side is Maximilian Negele, the report's co-author and an AI governance researcher at the Oxford Martin AI Governance Initiative. His camp's claim is that Europe's current response to frontier AI is ten to a hundred times too small and aimed at the wrong target, and that failing to compete at the frontier risks economic and political irrelevance. The Anthropic order lands as exhibit A: an early warning that access to the most advanced systems is a political variable, not a market one.

On the other side is Frederike Kaltheuner, an AI Now Institute advisor whose three-part Tech Policy Press series works bottom up rather than top down. Her argument is that even if Europe caught up on models tomorrow, the structural dependency would barely move. Startups like Lovable, in her reading, function as distribution channels for US models. Even independent European players like DeepL end up leaning on US hyperscaler infrastructure to serve customers. Sovereignty spending aimed at the frontier can quietly entrench that dependence rather than reduce it.

The honest caveat is that this is a podcast conversation between two advocates, not a settled body of research, and what the reporting doesn't give you is the counterfactual cost of substituting European alternatives at either the model or the infrastructure layer. Both guests are pitching Brussels, not shareholders, and the numbers behind the ten-to-a-hundred-times gap aren't broken down in the episode framing.

The framing itself is what makes it worth the hour. If the Anthropic order was the opening move rather than an outlier, European industrial policy has to answer two questions at once: who owns the frontier, and who owns the pipes and the distribution. Kaltheuner's contention is that only the second is fixable on European timelines. Watch which one the next round of AI industrial policy actually funds.