On 16 October, China’s Ministry of Commerce and Ministry of Science and Technology proposed new export controls on technologies used in AI models, AI agents and chip design. The draft adds 40 technologies to China’s list of technologies whose export is prohibited or restricted, 23 of them related to AI.
In plain terms, Chinese companies would need the government’s permission before transferring abroad four kinds of AI assets: leading AI models trained on sensitive Chinese data; advanced AI agents used in areas such as critical digital infrastructure, defence and government services; the know-how to design advanced processing and memory chips; and the technology needed to link AI chips into a single system. The last category is basically Huawei’s advanced AI computing hardware.
For the most advanced models, Beijing could restrict the transfer of the model itself, its source code and its design, and potentially stop offshore ownership or key employees moving abroad. Existing models that can be freely downloaded, even if very large, are exempted as long as they were not trained on sensitive data. In other words, the question is what happens with new advanced models.
The legislative proposal points to new frontier models no longer being freely downloaded abroad. Instead, they will need to be used on servers in China or on overseas servers run by Chinese firms. For example, a model like Kimi 3, one of the most advanced and popular Chinese AI models, would only be accessible on a Chinese server, inside or outside China.
In essence, while future open-source releases should get a fast track for automatic licensing, Beijing wants to retain the ability to limit the overseas transfer of its most advanced AI technologies and exercise greater control over how those technologies are shared and deployed.
Also read: Why China will not slow down its AI pace
Global stakes of China’s AI curbs
These rules, still subject to interpretation, could have two very different effects on the global market for AI models. If applied strictly, current users of Chinese open-weight models may worry about whether they will be able to obtain and run future versions on their own servers, as they may not want to rely on Chinese cloud services, especially for sensitive data. On the other hand, given their price advantage, global users may accept that their data is stored in a Chinese cloud.
Much remains uncertain about what Beijing wants to achieve, but the first impact on Chinese AI companies could be negative, as global users are now uncertain whether new versions of their models can be fully downloaded. And yet, if the world is already hooked on Chinese models, China might manage to move to a more profitable business model in which it sells not only its AI models but also its cloud services.
This is critically important for India. To start, the idea of using Chinese advanced models to reduce dependence on Chinese models looks like wishful thinking, as China is now creating its own technological dependencies. India is uniquely vulnerable to a reality where both the US and China limit access to their AI ecosystems, whether through hardware blockades or software restrictions. In fact, Indian developers and start-ups building on freely downloadable Chinese open-weight models may find that newer, frontier versions can only be used on Chinese servers, which is hardly an option for government services, defence or citizens’ personal data.
The difficulty with this reality is that India, like much of the world outside the US and China, cannot respond alone. It needs to advance partnerships with other countries that are also pursuing AI sovereignty, such as the EU, Japan, South Korea, the United Kingdom, or even the Gulf states.
The goal must be to pool computing capacity, jointly develop and openly share foundational model weights, and collectively secure access to advanced semiconductor chips. None of these countries can match the US or China on its own. Together, however, they could build an alternative that prevents the world’s most advanced AI from becoming a geopolitical lever that Washington or Beijing can pull at will.
Alicia García Herrero is Chief Economist for Asia Pacific and the Middle East at Natixis, and Senior Fellow at Bruegel. She tweets @Aligarciaherrer. Views are personal.
(Edited by Prashant Dixit)