Beijing weighs curbs on overseas access to China's advanced AI models

Chinese authorities have held meetings with leading technology companies over the past month to discuss potentially restricting overseas access to China's most advanced artificial intelligence (AI) models, including those yet to be released, three people familiar with the discussions said.

 

The talks follow a series of steps by Beijing to keep homegrown AI technology within the country and underscore how China, like the US, is increasingly treating cutting-edge AI as a critical national asset requiring tighter controls.

 

Companies that attended the meetings included Alibaba, ByteDance and startup Z.ai, the sources said. They were not authorised to speak to the media and requested anonymity.

 

 

Since the emergence of DeepSeek's R1 model last year, Chinese AI models have made significant inroads globally because of their low costs and improving capabilities. Any move by Beijing to restrict access to these products could have implications for global AI markets, potentially increasing costs for many businesses.

 

Tougher penalties for AI theft discussed

 

At the meetings, led by China's Ministry of Commerce, participants discussed imposing restrictions on the country's most advanced AI models, including both closed-source and more open versions, according to two of the sources.

 

Officials also discussed making the leak or theft of proprietary AI technology an offence under China's stringent national security law, one source said.

 

The officials further raised the possibility of introducing new measures to restrict who can invest in domestic AI startups, the source added.

 

The scope of the proposed restrictions remains under discussion and may apply only to future AI models, two sources said. It remains unclear when, or whether, the measures will be implemented.

 

China's Ministry of Commerce, which oversees export regulations, and the National Development and Reform Commission, whose officials also attended the meetings, did not respond to Reuters' requests for comment.

 

Alibaba, ByteDance and Z.ai also did not respond to Reuters' queries.

 

All three companies offer multiple AI models, including both closed-source and open-weight systems that users can download, run and customise.

 

Alibaba's Qwen and ByteDance's Doubao are among China's most widely used AI models. Z.ai has recently attracted attention in Silicon Valley as its GLM-5.2 model approaches the capabilities of leading US offerings at a fraction of the cost.

 

AI models and national security

 

US President Donald Trump's administration has also expressed concerns about the national security implications of AI, particularly the potential misuse of advanced American AI products by military or intelligence organisations in China, Russia and other countries of concern.

 

In June, the US ordered that foreign nationals should not have access to Anthropic's most advanced Fable and Mythos models, prompting the company to disable those models globally because it could not verify users' nationality in real time.

 

Export controls on Fable, designed for the general public, have since been lifted after new safeguards were introduced. Mythos, developed for cybersecurity professionals, remains available only to selected trusted US organisations.

 

Some US AI experts have also argued that the United States should regulate the use of Chinese AI models.

 

Concerns over Mythos

 

According to two of the sources, Chinese authorities are particularly concerned that Mythos could be used to identify software vulnerabilities and that Washington might deploy the model against Chinese interests.

 

These concerns echo comments made by Chinese state media and Zhou Hongyi, founder of cybersecurity firm 360, who has said China should develop its own version of Mythos.

 

China has introduced several measures this year to protect its domestic AI industry.

 

In April, the country's state planner ordered Meta to unwind its $2 billion acquisition of Chinese-founded AI startup Manus. In early June, authorities introduced sweeping new rules tightening oversight of overseas deals involving Chinese investors, technology, data and national security.

 

China has also launched investigations into Manus and other domestic AI startups that relocated overseas to determine whether they violated export control laws, according to two of the sources and a third person.

 

Manus did not respond to requests for comment.

 

Reuters could not determine how any new restrictions on overseas access to Chinese AI models would be implemented.

 

However, a possible framework emerged from a May roundtable of Chinese legal experts on open-source AI regulation. According to a summary published in an official journal of China's Supreme People's Court, participants proposed a tiered system under which basic open-source tools would require only simple filings, more advanced technologies would undergo security reviews, and the most sensitive frontier AI models would either be barred from public release or restricted to domestic use.