Key points
- Draft rules cover 23 AI technologies, including large models and chip designs.
- Open-source code and app-based access remain exempt from these controls.
- A temporary export registration rule took effect on 16 October.
China’s commerce and science ministries proposed adding 23 AI technologies to the country’s export control list. The draft includes large language models, AI agents, chip designs, and computing clusters. This move affects organizations relying on Chinese AI infrastructure or hardware.
By the numbers
- 23 AI technologies targeted by the proposed export controls.
- 40 total items in the broader proposal, including solar and biomedicine.
- 15 November deadline for public comment on the draft rules.
What the figures mean
The proposal covers models with at least 2 trillion parameters or those ranking in the top 10 globally. Bloomberg reported that Moonshot AI’s Kimi K3 and Alibaba’s Qwen3.8-Max meet this threshold. The draft also limits design know-how for AI chips made at 14 nanometres or below, or built using 2.5D or 3D stacking. Some memory chip designs are included.
Open-source code and published research are exempt. Chinese firms can still serve overseas users through apps, websites, and APIs. Running a covered model on servers abroad counts as an export, though firms using their own cloud servers may get automatic licenses. The ministry stated the rules aim to prevent misuse and ensure human control.
What should you do?
Managed service providers should audit their supply chains for dependencies on Chinese AI models or hardware. Check contracts for clauses related to export compliance and licensing. Monitor regulatory updates closely, as rules may change during the public comment period. Ensure internal teams understand the distinction between exempt open-source tools and controlled proprietary models.
Operational impact
Teams usually review vendor contracts to identify potential export control risks. Check whether current AI services rely on controlled technologies. Inform legal and compliance departments about new registration requirements. Assess if any overseas deployments might trigger export classifications. Coordinate with procurement to evaluate alternative suppliers if needed. Ensure documentation supports license applications for automatic approvals.
Background: Large language models
The difference between large language models and small language models lies in their parameter counts, training data volumes, and memory footprints. Large models handle complex, open-ended reasoning across many domains, while small models execute narrow tasks efficiently on local hardware with predictable latency and lower resource overhead.
What to do and how to stay safe: China
- Audit all AI models and hardware from Chinese vendors for compliance risks.
- Review service agreements for export control clauses and licensing terms.
- Monitor official announcements for updates to the draft regulations.
- Document current deployments to facilitate potential license applications.
Step-by-step guide: Shadow AI Explained: 8 Questions Answered for Secure Operations
General security guidance from the BitShield News newsroom. It is not confirmed advice from the organisations named in this story.
Frequently asked questions
Are open-source AI models exempt from the new controls?
Yes, the draft proposal explicitly exempts open-source code and published research from export controls.
When can the public comment on the proposed rules?
The public can submit comments on the draft proposal until 15 November.
Does running a model on overseas servers count as an export?
Yes, running a covered model on servers abroad counts as an export, though automatic licenses may apply for own cloud servers.
