China Plans to Restrict Export of Top AI Models; Alibaba, ByteDance Involved in Discussions
N.R. Finch
China's commerce ministry has held meetings over the past month with Alibaba, ByteDance and others to discuss restricting overseas access to its most advanced AI models; if enacted, foreign businesses relying on cheap Chinese AI face a direct cost shock.
What exactly is being discussed?
The talks are led by China's Ministry of Commerce, with officials from the National Development and Reform Commission also present.
The scope covers the most advanced AI models — both closed-source and open-weight versions, including models not yet released.
Participants include Alibaba, ByteDance, and startup Zhipu (Z.ai).
Are the restrictions settled?
Sources say the exact scope is still under discussion and may apply only to future model releases.
It remains unclear when — or whether — formal measures will land.
In plain terms = this is still at the "meeting to set direction" stage, not a done deal.
Could AI leaks become a criminal offence?
Officials raised the idea of classifying leaks or theft of proprietary AI technology as criminal violations of China's national-security law.
They also floated new restrictions on funding sources for Chinese AI startups.
This means → Beijing is looking beyond export curbs — it wants to control how technology leaks out and where the money comes in.
Why now?
Since DeepSeek R1 emerged, Chinese AI models have expanded globally on the back of low cost and rapidly improving capability.
Alibaba's Qwen and ByteDance's Doubao are the most widely used models domestically; Zhipu's GLM-5.2 recently drew attention in Silicon Valley, with capability close to top US products at a fraction of the cost.
This reflects a shift: Chinese AI is no longer a follower — and it is precisely that global competitiveness that makes "should we restrict exports?" a live question.
Are China and the US doing the same thing?
The US has imposed its own national-security restrictions on AI access. In June, the Trump administration barred foreign nationals from accessing Anthropic's most advanced Fable and Mythos models.
Fable has since been restored under new safeguards, but Mythos remains available only to select "trusted" US institutions.
In plain terms = the two countries are mirror images — both locking down their strongest AI on national-security grounds, differing only in timing and tightness.
What does this mean for overseas users?
If restrictions land, foreign businesses that turned to Chinese models for their low pricing will face a cost shock.
Beijing has already moved to protect domestic AI this year: in April the NDRC ordered Meta to unwind its roughly $2 billion acquisition of Manus; in June new rules tightened controls on overseas deals involving Chinese investors, technology and data.
This means → whether these curbs materialise — and how broadly — is the key variable for anyone tracking China's AI globalisation strategy.
Content is for reference only, not financial advice.