Skip to content
Menu
Menu

China Proposes Export Limits on 23 AI Technologies, Including Its Largest Models

Starting Oct. 16, Chinese companies must register with the government before they sell or license any of the 40 technologies to a foreign business, even though the list is not final.

 

China’s Ministry of Commerce and Ministry of Science and Technology proposed adding 40 technologies, 23 of them in AI, to China’s official list of technologies that companies cannot export or can export only with a government license. The AI items include China’s largest AI models, AI that finds security flaws in software, and AI chip designs. The other 17 cover solar panel manufacturing and biomedicine.

As of October 16, a Chinese company must register with the Commerce Ministry before selling or licensing any of the 40 technologies to a foreign business. Once the list is final, 14 of the 40 could not be sold or licensed abroad at all, and the other 26 would need a license.

Foreign businesses could still use a Chinese AI model through an app or website, or download one already released for free, without the Chinese company registering or obtaining a license.

China would require a license to export its largest AI models and AI that finds security flaws

Of the 23 AI technologies, 17 would need a license, and 6 would be banned. Those needing a license include:

  • Large AI models. A Chinese company would need a license to hand over a model’s core files, meaning the values it learned in training, its source code, or its design. The rule would apply to any model with 2 trillion or more of those values, called parameters; one trained with at least 10^26 calculations; or one ranked in the world’s top 10 by major testing groups or large AI platforms. The list names no models. AP reports that Moonshot’s Kimi K3 and Alibaba’s Qwen 3.8-Max are over 2 trillion parameters.
  • AI that finds security flaws. Examples include AI that finds and exploits flaws 10 times faster than a senior security expert, or finds five or more previously unknown serious flaws in a major operating system in one run.
  • Advanced AI agents, meaning AI systems that carry out tasks on their own. Examples include agents that can break a request into 20 or more steps or use 200 or more outside software tools.
  • AI chip designs and the technology that links more than 10,000 AI chips into one computing system.

The banned AI items include software that uses AI to write malicious code and keep rewriting it so antivirus programs miss it. The ban also covers AI agents built mainly for critical infrastructure, defense, and core government systems, as well as large models trained on what Chinese law calls “core data.” The other 8 banned items are in solar and biomedicine.

The Commerce Ministry said the malicious-code tools could be used to attack critical infrastructure, defense, and government systems. It said the controls are meant to prevent misuse and ensure AI “always remains under human control.”

China would not require a license for apps or free downloads

The ministries said technology already made public through open-source release, where a company publishes a model’s files for anyone to download, or through academic papers is not covered.

The Commerce Ministry said a Chinese company would not need a license to let foreign customers use a restricted model through an app, a website, or software they log in to online, because the model’s files stay in China. Running the model on servers abroad would require a license, which the company would automatically receive if it owns the servers. Open-source releases and joint research with foreign partners would also be approved automatically.

The ministry said most AI technology would not need a license.

Companies must register now, and public comments close Nov. 15

Until the final list takes effect, Chinese companies must register on the Commerce Ministry’s online system before selling or licensing any of the 40 technologies to a foreign business. This includes deals that began before Oct. 16 but are not yet complete. The requirement covers selling a technology outright to a foreign buyer, licensing it for use abroad, and a foreign buyer taking control of the Chinese company that owns it.

The company must provide the buyer’s details, the destination country, the technology, and the deal’s expected value. The announcement does not say what happens to companies that do not register.

The public can comment until Nov. 15. Neither ministry gave a date for the final list. The Commerce Ministry said it has updated the list many times since 2001 and that this update is not aimed at any specific country, region, or company.

The proposal follows the U.S.-China September agreement to hold formal government talks on AI risks, with a meeting planned in Shenzhen.

Clayton Rifkind

Clayton Rifkind is the Founder and Senior Editor of AI Risk Today. He also advises on business development for ESG Today, a leading source of ESG investment news and research for institutional investors and corporate leaders. He has 20+ years of experience in B2B technology, leading strategy and execution of go-to-market plans across software, enterprise platforms, and mobile applications. He founded two consultancies advising startups and Fortune 1000 companies, including Autodesk, Intel, and Microsoft. He began his career in the San Francisco advertising scene working with brands such as Hewlett-Packard, Intel, Microsoft, Symantec, and Wells Fargo. Clayton launched AI Risk Today in 2025 after two decades of watching enterprises adopt transformative technologies, and seeing how often risk, governance, and compliance considerations lagged behind. His reporting draws on primary sources including regulatory filings, court documents, and official announcements, with a focus on what AI developments mean for the executives accountable for managing them. Reach him at Reach him at [email protected] or on LinkedIn.

Essential AI Risk Intelligence

Daily insights on AI governance, regulation, and enterprise risk management. Trusted by Chief Risk Officers and compliance leaders globally.

By subscribing, you agree to receive our daily newsletter. Unsubscribe anytime.

Advertise with AI RIsk Today, Today!