Published: October 17, 2026 · Last updated: October 17, 2026
China has proposed adding AI model weights, training data, and chip design tools to its export control list, a draft rule that would let Beijing require a license before any of China’s competitive open-weight AI models leave the country.

For years, the story of AI export controls ran in one direction: the United States restricting what chips and chipmaking tools could reach China. This week China proposed flipping part of that script. Its Ministry of Commerce published a draft update to its catalogue of technologies subject to export restrictions, and for the first time it explicitly lists AI model weights, training data pipelines, and chip design tools among the 23 technologies it wants tighter control over.
The draft, first reported by the Financial Times and confirmed by multiple outlets since, does not ban these exports outright. It adds them to a licensing regime, meaning Chinese companies would need government approval before selling or transferring the listed technologies abroad. That is a meaningfully different posture than an outright ban, but it hands Beijing a lever it did not clearly have before: the ability to slow, condition, or block the export of open-weight Chinese AI models and the tools used to build them.
The context matters here. Chinese open-weight models, most visibly from DeepSeek and a handful of other labs, have spent the last two years becoming genuinely competitive with top American systems, and they are free for anyone outside China to download, modify, and build on. That openness has been one of China’s few unambiguous wins in the broader AI race, letting Chinese labs build global developer mindshare even while facing chip restrictions from Washington. A licensing requirement on exporting model weights would not stop that openness, but it would give Beijing a formal checkpoint it currently lacks.
There is also a defensive read on the timing. Washington has spent the past two years tightening its own chip export rules aimed at China, treating advanced semiconductor technology as a national security asset worth controlling at the border. China’s draft rule borrows the same legal logic and points it at a category the United States does not currently restrict in the same way, the AI models themselves, not just the hardware underneath them. Framed that way, this looks less like China closing itself off and more like China building the same kind of export lever the United States already has, in case it needs to use it.
The draft is open for public comment before it becomes binding, which is standard practice for Chinese regulatory changes and typically takes weeks rather than days. Industry groups and foreign governments are expected to weigh in, and the final list of covered technologies could shrink or grow before it takes effect.
What it means in practice, if adopted largely as proposed, is a new layer of friction for any foreign company, lab, or research group that currently relies on downloading Chinese open-weight models or buying Chinese AI tooling. It would not necessarily stop that flow, but it would make it something Beijing can switch off selectively, the same way the United States can with its own export licensing regime.
Related: a chip startup just raised $125 million betting on the hardware side of this same AI buildout and the banks financing that buildout are starting to get cold feet.
Bottom Line: China is not shutting the door on its AI exports, it is installing one, and the ability to open or close it at will is the point. Watch the public comment period closely, since the final list of covered technologies will say a lot about how far Beijing actually intends to go.
Follow Teck Hustlers: Facebook · X · Instagram · LinkedIn · TikTok · Pinterest



