China's Ministry of Commerce is drafting AI export controls that go beyond hardware to restrict the transfer of model weights, training pipelines, and even research talent to foreign entities. Three companies have already received administrative warnings, and Chinese nationals working on frontier systems could be classified as holders of 'strategic technical knowledge' — the same designation applied to nuclear and missile specialists.
The US-China AI fight has a new front. For two years the battle was about hardware — Nvidia GPUs, ASML lithography machines, advanced packaging. But Chinese regulators are now drafting export controls that go far beyond chips. They want to restrict the flow of AI talent, model weights, training data, and the intellectual property that makes frontier AI systems work.
The shift signals that Beijing views AI as a strategic asset requiring the same level of protection as nuclear technology or advanced weapons systems. It also changes the calculus for every AI company operating across both markets.
What's Being Proposed #
China's Ministry of Commerce is leading the effort, with input from domestic AI firms including Alibaba, ByteDance, and Baidu. The draft rules would require export licenses for transferring model weights, training pipelines, and inference infrastructure developed by Chinese companies to foreign entities.
The scope is deliberately broad. It covers not just completed models but the entire development stack — training code, data curation methods, fine-tuning recipes, and evaluation frameworks. The intent is to prevent foreign competitors from replicating Chinese AI advances using Chinese-developed intellectual property.
Three companies have already received administrative warnings in the first week of enforcement, according to Ministry guidance published July 9. The rules are operational, not theoretical.
The Talent Dimension #
The more novel element is the proposed restriction on talent flows. The draft rules would classify certain AI researchers and engineers working on frontier systems as holders of "strategic technical knowledge" — a designation that already applies to nuclear and missile technology specialists under Chinese law.
Under this framework, Chinese nationals with deep expertise in frontier model development could face restrictions on working for foreign companies, attending overseas conferences, or publishing research deemed sensitive. Several AI firms are reportedly reviewing their international collaboration agreements in response.
This marks a significant escalation. Previous US-China AI tensions focused on hardware supply chains and app bans. Restricting the movement of people and the code they write takes the conflict to a different level.
The Strategic Logic #
Beijing's calculation is clear. After watching the US restrict chip access, restrict investment in Chinese AI, and pressure allies to do the same, China is building its own fences. If the US treats AI as a domain of strategic competition, China will do the same — using the same tools.
The difference is that China's AI sector has reached a level where export controls actually matter. When DeepSeek was a scrappy underdog, nobody cared about restricting its model weights. Now that Chinese labs are producing frontier models that compete on benchmarks, Beijing has something to protect.
What Comes Next #
The draft rules are open for comment but enforcement is already starting through administrative channels. Companies operating AI labs across both the US and China face the most immediate compliance challenges — their researchers, code, and infrastructure now sit across a hardening border.
The EU AI Act's August 2 enforcement deadline and the US Senate's Great American AI Act are already reshaping the regulatory landscape. China's export control push adds a third major regime to track. For global AI companies, compliance is becoming as complex as the models themselves.
Q: What exactly would China's new export controls cover?
Q: Are these rules already in effect?
Q: What's the talent restriction angle?
Related Articles #
Get AI news in your inbox
Daily digest of what matters in AI.
Key Terms Explained #
Evaluation The process of measuring how well an AI model performs on its intended task.
Fine-Tuning The process of taking a pre-trained model and continuing to train it on a smaller, specific dataset to adapt it for a particular task or domain.
Inference Running a trained model to make predictions on new data.
NVIDIA The dominant provider of AI hardware.