China Discusses Restricting Overseas Access to Advanced AI Models

Chinese authorities have held discussions with some of the country's biggest technology companies over the past month on restricting overseas access to China's most advanced artificial intelligence (AI) models, including those that have not yet been released, according to three Reuters sources.

The talks, led by China's Ministry of Commerce, reflect Beijing's growing effort to treat cutting-edge AI as a strategic national asset, similar to the United States, amid intensifying global competition over the technology.

Officials discussed potential restrictions on both closed-source and open-weight AI models, with the scope of the proposed measures still under consideration.

According to two sources, the curbs may apply only to future AI models, and it remains unclear if or when they will be implemented.

Representatives from major Chinese technology firms, including Alibaba, ByteDance and AI startup Z.ai, attended the meetings, the Reuters sources said.

The discussions also covered strengthening legal protections for AI technology.

According to one source, officials proposed making any leak or theft of proprietary AI technology an offence under China's national security law.

They also explored imposing tighter restrictions on who can invest in domestic AI startups.

The move comes as Chinese AI models rapidly gain global traction.

Since the launch of DeepSeek's R1 model last year, Chinese developers have attracted international attention by offering increasingly capable AI systems at significantly lower costs than many Western rivals.

Industry observers say any restrictions on overseas access could have ripple effects across the global AI market by limiting availability of low-cost models.

Alibaba's Qwen and ByteDance's Doubao are among China's most widely used AI models, while startup Z.ai has drawn attention in Silicon Valley after its GLM-5.2 model demonstrated capabilities approaching leading U.S. AI systems at a fraction of the cost.

The discussions come amid growing concerns in both Beijing and Washington over AI's national security implications.

Beijing has introduced several measures this year to strengthen oversight of its AI sector.

In June, authorities tightened regulations governing overseas deals involving Chinese investors, technology, data and national security.

Officials have also launched investigations into Manus and other Chinese AI startups that relocated abroad to determine whether they violated export control rules, according to three sources.

The US' Trump administration has also tightened access to advanced American AI systems.

In June, it barred foreign nationals from accessing Anthropic's most advanced Fable and Mythos models, prompting the company to temporarily disable the models globally because users' nationalities could not be verified in real time.

While export controls on Fable were later eased after additional safeguards were introduced, Mythos remains available only to select "trusted" US organisations.

According to two sources, Chinese officials are particularly concerned that Mythos could be used to identify software vulnerabilities and potentially be deployed against Chinese interests.

Those concerns echo warnings from Chinese state media and Zhou Hongyi, founder of cybersecurity firm 360, who has argued that China should develop its own equivalent of Mythos.