China won’t stop its AI from going global, but it wants control over how it travels

A state media-affiliated account in China said on October 22 that DeepSeek and MiniMax had completed the registration process for releasing artificial intelligence (AI) models openly and licensing them for overseas use.

The announcement came days after Beijing proposed new export controls covering 23 AI-related technologies, including large language models (LLMs), AI agents and computing clusters.

People and developers around the world, however, were already using DeepSeek. So what does China’s new approval for global use of its AI models actually change?


The proposed rules have raised concerns that China could make it harder for its AI companies to take their models overseas, at a time when developers globally were turning to Chinese models as a cheaper alternative to US offerings.

What has changed?

The approval does not mean DeepSeek is now allowed to operate internationally for the first time. Its models were already available to users outside China.

The difference lies between allowing people to use an AI service and allowing the underlying technology to be transferred, licensed or deployed overseas.

In an explanation published by Yuyuan Tantian, a social media account affiliated with state broadcaster China Central Television (CCTV), Chinese authorities drew a distinction between exporting AI technology and providing access to AI services.

According to the account, overseas access through APIs, websites and mobile apps would not be covered by the proposed export restrictions, even when those services use models subject to the controls.

In other words, using a Chinese AI model through an online service is different from transferring the underlying technology or deploying it on servers overseas. The latter is where Beijing wants greater oversight.

Why tighten oversight now?

Chinese AI companies have gained global traction by offering relatively affordable models that developers can download, customise and use in their own applications. This has made Chinese models an alternative to proprietary systems from US companies such as OpenAI and Anthropic.

But as AI becomes more strategically important, Beijing wants greater control over how the underlying technology moves across borders.

The draft rules were released on October 16 by China’s Ministry of Commerce and Ministry of Science and Technology, and are open for public comment until November 15. According to Bloomberg, they cover 23 AI-related technologies and seek to limit transfers of tools that China considers important to national security.

The broader draft revision of the export control list also covers technologies in other sectors, including biopharmaceuticals and solar energy.

Chinese state media has described the policy as an effort to set clearer boundaries for technology exports rather than stop them altogether. The government has also cited concerns about misuse and the need to keep AI “under human control.”

Will this affect people globally?

For now, there is little indication that ordinary users will lose access to these models. Yuyuan Tantian has said most AI technologies will remain freely exportable, while the draft rules offer automatic licensing for open-source releases.

The proposed rules do, however, change what global developers need to watch. Accessing a model through an online service may remain straightforward, while exporting, licensing or deploying the underlying technology overseas could involve additional regulatory requirements.

This matters because Chinese open-weight models have become an important resource for developers and research institutions looking for alternatives to US AI systems.

The bigger takeaway is that Beijing is not necessarily trying to stop its AI models from spreading globally. It wants to retain control over how that technology is exported.