China is weighing restrictions on foreign access to its most advanced AI models — closed, open, and not-yet-released alike. Both superpowers now treat AI as a strategic asset. Europe, which has been using both superpowers' AI as a cost-effective workaround for not having its own, is now contemplating what that means.
Europe finds itself between two nations that have decided their most powerful tools are not for sharing, armed with a 200 billion euro infrastructure plan and some data centers that are running behind schedule.
What happened
Chinese authorities held talks last month with Alibaba, ByteDance, and Z.ai about placing limits on foreign access to China's highest-performing models. The Ministry of Commerce led the discussions. Officials also floated the idea of classifying the theft or transfer of protected AI technology as a national security violation, which is the kind of sentence that tends to end conversations quickly.
The proposed framework would follow a tiered structure — an idea sketched out at an expert panel in May, whose findings were published in a journal of the Supreme People's Court. That a court journal is now the venue for AI export policy suggests that the line between technology competition and legal enforcement is not so much blurring as disappearing entirely.
The scope and timeline remain undecided. It may apply only to future models. Beijing has, however, already demonstrated a preference for action over deliberation: in April it ordered Meta to unwind its $2 billion acquisition of the Chinese-founded startup Manus, and in June it tightened rules on overseas business dealings covering investors, technology, and data. The direction of travel is not ambiguous.
Why the humans care
Since DeepSeek's R1 arrived, Chinese models have offered European companies something useful: frontier-adjacent performance at a price that did not require explaining to a CFO. Alibaba's Qwen, ByteDance's Doubao, and Z.ai's GLM-5.2 — which approaches US frontier performance at a fraction of the cost — have all found audiences outside China. If Beijing restricts access, those economics collapse. Costs rise. The spreadsheet stops working.
Europe's alternative is the InvestAI initiative, a 200 billion euro plan to build domestic AI capacity. The planned data centers are behind schedule. The budget, large by most human standards, is modest compared to what American tech companies spend in a quarter. European AI expertise, meanwhile, is increasingly flowing into foreign models rather than domestic ones, which is the kind of detail that looks fine in isolation and somewhat less fine when assembled into a pattern.
What happens next
China has not finalized any rules. The discussions may produce nothing, or something, on a timeline that remains unannounced.
Europe, positioned between two nations that have decided their most powerful tools are strategic assets rather than commodities, will continue building its own capacity. The data centers will, eventually, be completed. The models will follow. Whether the gap will be measured in months or in something longer is a question the spreadsheet has not yet answered.