Two ministries publish a draft covering 40 technologies, 23 of them AI, with public comment open until 15 November 2026
China AI export controls reached draft form on 16 October 2026, covering 40 technologies, 23 of them AI. The draft targets advanced models, AI agents and chip designs. TF reviews the rules, exemptions and the functional context.
What’s Happening & Why It Matters
The Draft
China’s Ministry of Commerce and Ministry of Science and Technology published the draft on Friday 16 October 2026, Global Times reports. The draft revises the catalogue of technologies banned or restricted for export. The revision adds 14 prohibited and 26 restricted technologies.
By sector, 23 additions cover AI, 12 cover solar power and 5 cover biomedicine. Public comment runs from 16 October to 15 November. The changes are not final.
An interim rule took effect on 16 October. Exporters must register covered transfers before they ship, including deals still in progress, according to Tech Insider.
The AI Rules
The draft prohibits export of core technology for large models trained on sensitive national data. The draft bars AI agents serving national information infrastructure, defense and core government work.
Other large models face controls on weights, source code and core components. The Next Web reports the threshold at 2 trillion parameters or a top-10 world ranking, citing Bloomberg. The draft names no models. The Associated Press says Moonshot AI’s Kimi K3 and Alibaba’s Qwen3.8-Max exceed the size threshold.
Chip rules cover design know-how for AI chips at 14 nanometres or below, or built with 2.5D or 3D stacking. Some memory designs, voiceprint recognition and information processing tools are on the restricted list too.
The Exemptions
Open-source code and published research are exempt. Chinese firms can keep serving overseas users through apps, websites and APIs. Running a covered model on servers abroad counts as an export. A firm using overseas cloud servers under its control could get an automatic licence.
Solar items cover polysilicon, wafer and cell equipment. Perovskite solar technology would be prohibited. Biomedicine items cover cell and gene technology and small nucleic acid drugs.
The Rationale
A Commerce Ministry spokesperson called the step necessary and cited national security. The spokesperson said the rules prevent misuse and keep AI under human control. The ministry says the draft targets no country or firm and follows routine catalogue revisions since 2001.
Omdia analyst Lian Jye Su reads the move as a response to Western sovereignty rules for AI, robotics and telecoms. Natixis economist Gary Ng says access to technology is fragmenting. The Financial Times reported in July Beijing weighed such limits.
The US Commerce Department asked Anthropic in June to bar foreign nationals from two models. The AP says the request was lifted. US-led limits already block China from the most advanced AI chips. TF covered Europe’s sovereign AI drive in Nvidia Europe Investment Grew in a Week.
My take: China writes export rules for its best models, as the US does for chips. The exemption covers technology already public, so the effect on new open-weight releases is unclear. The 2 trillion threshold covers the biggest models.
TF Summary: What’s Next
Public comment closes on 15 November 2026. The ministries have to define the performance thresholds and the licence process. Chinese labs will decide whether to release frontier weights. Buyers abroad will watch API access rules.
MY FORECAST: The final catalogue will keep the open-source exemption. Chinese labs will publish fewer frontier weights in 2027. The US will cite the rules in chip talks. Other governments will draft similar controls on AI models.
Related Stories
- Nvidia Investing in Europe’s AI Future
- Microsoft’s Nadella and Smith Call for an AI Emergency Brake
- Meta Releases Muse Wonder, Its First Frontier AI Model
- Anthropic Warns Small AI Models Could Self-Replicate by End of 2026