China mulls export curbs on AI models; Europe risks access

China is weighing restrictions on foreign access to its most advanced AI models after meetings last month with leading Chinese tech firms. The talks, led by the Ministry of Commerce, included Alibaba, ByteDance and the startup Z.ai and covered limits on high-performance systems, both closed and open models, as well as unreleased models.

What is China proposing and who was involved?

China explored a tiered approach to controlling model access and ownership, with proposals that would subject advanced technologies to security reviews and possibly bar the most sensitive systems from foreign use. Officials discussed placing theft or transfer of protected AI technology under China’s national security law and tightening who can fund domestic AI startups. An expert panel’s summary suggested the "most sensitive frontier models would either not be released publicly at all or be limited to domestic use." Reuters, cited in the reporting, named Alibaba, ByteDance and Z.ai as attendees and said the talks were led by the Ministry of Commerce.

How would this affect Europe’s AI access and investment plans?

Europe risks losing a low-cost alternative to US models because China’s suggested limits would make Chinese frontier models unreliable for foreign users. Chinese models such as Alibaba’s Qwen, ByteDance’s Doubao and Z.ai’s GLM-5.2 have gained ground by offering lower-cost options; GLM-5.2 is described as approaching US frontier performance at a fraction of the cost. The EU has launched InvestAI, a roughly 200 billion euro initiative announced in early 2025 that includes 150 billion euros from a private "AI Champions" initiative and 50 billion euros in additional European investment. Twenty billion euros are earmarked for up to five AI Gigafactories. The EU Council adopted the legal framework for InvestAI in January 2026, but the formal tender for the Gigafactories has been pushed back to summer 2026 and construction is scheduled to begin in 2027. Meanwhile, four US tech companies could spend a combined $700 billion on AI-driven investments in 2026 alone, a sum the reporting says is nearly double their prior year spending and roughly three times InvestAI’s entire multi-year vision.

How does this compare with recent US policy?

Washington has already restricted access to some advanced models. In June the US barred foreign nationals from accessing Anthropic’s most advanced models, Fable and Mythos. Because nationality could not be verified in real time, Anthropic initially shut those models worldwide; export controls on Fable were later lifted after new safeguards, while Mythos remains restricted to a handful of trusted US organizations and vetted partners. Chinese officials voiced concern that tools like Mythos could be used to exploit software vulnerabilities against Chinese interests, and local voices, including Zhou Hongyi of security firm 360, have called for Chinese equivalents to Mythos.

Why it matters

Europe relies heavily on external AI providers and on the free flow of talent and expertise. The EU depends on foreign providers for more than 80 percent of digital products, services and infrastructure, and the single European market of roughly 450 million consumers plus regulatory influence have so far preserved access in many cases. That balance could shift if both Washington and Beijing treat frontier models as strategic assets. Europe already faces a chronic funding gap: more than four out of five major EU funding rounds have a foreign lead investor, compared with 14 percent in San Francisco. AI deepens the leakage by allowing specialist expertise to be bought and fed into foreign models directly, accelerating a transfer of intellectual capital even when companies remain physically in Europe.

What to watch

Watch whether China finalizes any export or investment curbs and the scope they set for "advanced" or "frontier" models; look for the formal text and implementation timeline. Also track the InvestAI Gigafactory tender expected in summer 2026 and whether the first sites begin construction in 2027, and compare announced EU and US AI spending over 2026 for shifts in relative capacity and market access.