Technology / Politics

Add as a preferred source on Google

China Proposes Export Controls on 23 AI Technologies in Wider Draft

The proposal adds 40 technologies across AI, biomedicine and photovoltaics; a separate registration rule took effect Oct. 16.

Updated 2 min read
Staff enter China's Ministry of Commerce headquarters in Beijing beneath an overcast sky. / EconoTimes.ai
Staff enter China's Ministry of Commerce headquarters in Beijing beneath an overcast sky. / EconoTimes.ai

China’s Ministry of Commerce (MOFCOM) and Ministry of Science and Technology (MOST) have proposed new export controls covering 23 artificial intelligence technologies, placing them in a broader revision that also requires advance registration.

The proposal, announced Oct. 16, 2026, would add 40 technologies across artificial intelligence, biomedicine and photovoltaics to the export catalogue. The AI entries include large models, agents, computing clusters and information processing.

MOFCOM distinguishes transferring restricted large-model technology overseas from providing access to a model. The ministry says overseas localization of restricted large-model technology falls under export administration, while access to a restricted large model through APIs, apps, websites or software-as-a-service platforms does not constitute exporting the technology. It also excludes lightweight on-device deployment from that definition. A company’s own overseas cloud deployment may qualify for automatic licensing or other facilitation.

A separate temporary rule took effect on Oct. 16. Announcement No. 48 requires exporters to register before sending covered technologies abroad while the catalogue revision is pending. The requirement also reaches some unfinished exports already in progress, such as transfers of technology ownership to foreign entities, licenses for overseas use and changes that give a foreign entity ownership or actual control.

MOFCOM says technology already in the public domain through open-source releases or academic publications would not be covered by the proposed controls. It cites concern that technology for using AI to generate or modify malicious code could be used to attack critical information infrastructure, defense-related systems and core government systems. The ministry describes the catalogue revision as routine, says similar revisions have taken place multiple times since 2001 and says the proposal is not directed at a particular country, region or entity.

Reuters, citing a person familiar with the discussions, reported that China could finalize the proposals before 2027 and publish implementation guidance afterward. The official notice does not set a date for the revised catalogue to take effect. Public comments are due by Nov. 15, 2026.

Related articles

View all