China Proposes Export Controls on 23 AI Technologies, Including Frontier Language Models
China's commerce and science ministries have published a draft proposal to restrict exports of 23 AI technologies — including large language models above two trillion parameters and advanced chip designs — as part of a broader 40-item list, with public comments open until 15 November.
China's Ministry of Commerce and Ministry of Science and Technology jointly released a draft proposal to add 40 technologies to the country's export-control catalogue. Of those, 23 fall under artificial intelligence; the rest cover 12 solar-energy items and five in biomedicine, including gene therapy. The public comment period closes on 15 November.
For AI, the draft would restrict exports of large language models — covering their weights, source code, and other core components — at a threshold of at least two trillion parameters, or a ranking in the world's top ten on performance and safety benchmarks. Bloomberg and the Associated Press identified Moonshot AI's Kimi K3 and Alibaba's Qwen3.8-Max as models that would fall above that size; the draft itself does not name any specific model.
The proposal would also limit design know-how for AI chips manufactured at 14 nanometres or below, or built using 2.5D or 3D chip-stacking methods, along with certain memory chip designs. AI agents and computing clusters are also on the list.
Open-source code and published research are explicitly exempt, and Chinese companies may continue serving overseas users through apps, websites, and APIs. Running a covered model on foreign servers would count as an export, though firms using their own cloud servers abroad could receive an automatic licence.
A temporary export-registration requirement, which also covers deals not yet finalised, took effect on 16 October. A ministry spokesperson said the rules aim to "prevent misuse and abuse" and "ensure that AI remains under human control," adding that the plan does not target any specific country or company.
