Beijing weighs curbs on overseas access to China's most advanced AI models

China’s Ministry of Commerce has held meetings over the past month with Alibaba, ByteDance and the startup Z.ai to discuss restricting overseas access to the country’s most advanced AI models, according to three people familiar with the discussions. The restrictions would extend to some models not yet released.

The talks point to a shift in China’s AI strategy. Chinese open-weight models have become a growing source of soft power abroad, and Beijing is now weighing whether its most capable systems should stay at home rather than spread freely to overseas users.

Proposals under discussion include a tiered system, the people said. Basic open-source tools would need only a simple filing; more advanced technologies would face security reviews; and the most sensitive frontier models would be barred from public release or restricted to domestic use. Officials also discussed classifying the leak or theft of proprietary AI technology as an offense under China’s national security law, along with new limits on who can fund domestic AI startups. A summary of the discussions was published in an official Supreme People’s Court journal.

The stakes are tied to how far Chinese models have already traveled. Open-weight systems such as Alibaba’s Qwen and ByteDance’s Doubao have won significant overseas adoption since DeepSeek released its R1 model, helped by low cost and improving capability. Curbs would put that reach in tension with control.

The reporting rests on unnamed sources, and the plans are unsettled. The scope of any restrictions remains undecided, and it is unclear whether or when they would take effect. China’s Ministry of Commerce and National Development and Reform Commission did not respond to requests for comment, nor did Alibaba, ByteDance or Z.ai.

Whether Beijing formalizes a tiered regime, and where it draws the line on frontier models, will show how it balances AI influence abroad against security control at home.