China has proposed adding 40 technologies to its national export control list, 23 of them in artificial intelligence, in a draft plan that would for the first time bring AI model weights and source code inside Beijing’s export-control perimeter. The China AI export controls were announced on October 16 by the Ministry of Commerce and the Ministry of Science and Technology, with public consultation open until November 15.
The move marks the most direct attempt yet by Beijing to control the software layer of artificial intelligence rather than just the hardware beneath it. Reuters reported that the proposed changes cover large language models, AI agents, chip designs and computing clusters, which it described as the three fundamental building blocks of China’s AI industry, underscoring why the China AI export controls are being watched so closely in Washington.
What the China AI export controls would cover
Across the full 40-item catalogue, 14 technologies would be prohibited from export outright and 26 would be restricted under licensing. The banned list includes AI-driven technology that can generate or mutate malicious code and AI models trained on sensitive Chinese national data. The restricted list covers AI-assisted vulnerability discovery as well as technologies connected to advanced models, agents and computing infrastructure. Analysts say the China AI export controls reflect a broader turn toward treating advanced computing as a strategic resource on par with rare earths and advanced chips. The Associated Press reported that models with at least two trillion parameters, or with specified globally leading capabilities, could fall into the restricted category.
Significantly, model weights and source code would be captured only when products exceed certain performance thresholds, details of which the published summaries do not quantify. The draft exempts access through APIs, websites, apps and software-as-a-service platforms, and leaves alone technologies already in the public domain through open-source releases or academic publications. The ministries have also put an interim pre-export registration requirement into force immediately, meaning the paperwork starts now even while the rules are still being finalised. The China AI export controls would also reach transfers of technology ownership, overseas licensing and moves of ownership to foreign entities, according to the draft, and exports already underway but not completed would fall within scope.
Why Beijing says the China AI export controls are needed
In a separate statement, the commerce ministry said the revision was proposed because applications of related technologies in sensitive fields are constantly expanding, and described the move as routine, stressing that it does not target any specific country, region or entity. A ministry spokesperson said the measures aim to ensure AI technology remains safe and controllable, and that “the vast majority” of AI-related technologies would remain freely exportable, calling the scope “precise and limited”.
The ministry also warned that technologies such as AI models used for malicious coding could be exploited to attack critical national computing infrastructure, defence installations and core government functions. “If acquired by non-state actors or malicious entities, these technologies could pose a grave threat to national security and common global security,” it said. Officials framed the China AI export controls as consistent with international norms on dual-use technology, arguing that governments everywhere are tightening oversight of capabilities that could be turned against critical infrastructure. The ministry has been consulting domestic AI and chip firms including Alibaba, ByteDance and Zhipu on the shape of the controls since the summer.
How the China AI export controls fit the wider tech race
Beyond AI, the draft covers five biomedicine technologies, including cell and gene therapy platforms, and 12 solar technologies covering polysilicon, wafers and cell manufacturing. The proposal lands amid an escalating technology rivalry: both Washington and Beijing have accused each other in recent months of misappropriating locally developed AI models, and the Wall Street Journal reported in July that Chinese officials had discussed with domestic AI labs how to protect their proprietary technology.
Chinese President Xi Jinping has called for tighter safeguards on AI, while U.S. President Donald Trump has urged more limited government regulation. For foreign companies, the China AI export controls create immediate compliance questions: the interim registration requirement means firms must now log covered transfers before proceeding, even though the final rules are still open for comment. Trade lawyers say the exclusion of open-source technology and API access will soften the blow, but the first-time inclusion of model weights marks a structural shift in how Beijing treats AI as a strategic asset. Industry groups and foreign firms operating in China will now be watching the consultation closely, since how broadly the final China AI export controls are drawn will determine how much of China’s AI stack can still move across borders.