China Reportedly Weighs Restricting Overseas Access To Advanced AI Models As Beijing Tightens National Security Controls

Chinese authorities are considering new restrictions on overseas access to the country’s most advanced artificial intelligence models, signaling Beijing’s determination to treat frontier AI technologies as strategic national assets amid intensifying technological competition with the United States.

According to people familiar with the discussions cited by Reuters, senior Chinese officials have held a series of meetings over the past month with leading technology companies, including Alibaba, ByteDance and AI startup Z.ai, to explore measures that could limit foreign access to China’s most sophisticated AI systems, including next-generation models that have yet to be released.

The discussions, led by China’s Ministry of Commerce and attended by officials from the National Development and Reform Commission (NDRC), represent Beijing’s latest effort to tighten oversight of an industry that has become central to economic competitiveness, military capability and national security.

If implemented, the restrictions could reshape the AI industry, particularly as Chinese models have emerged as increasingly credible alternatives to expensive U.S. offerings.

The deliberations show that China is adopting an approach similar to Washington’s by treating advanced AI as sensitive technology requiring government oversight. Policymakers are now categorizing sophisticated AI models alongside other strategic technologies such as advanced semiconductors and quantum computing.

According to two of the sources, officials discussed imposing restrictions on both proprietary closed-source models and open-weight systems that currently allow developers to download, modify, and deploy the underlying models. Authorities are also reportedly considering making the theft or unauthorized transfer of proprietary AI technology a criminal offence under China’s sweeping national security laws.

Another proposal under discussion would tighten scrutiny of investors funding domestic AI startups, reflecting concerns that foreign capital could become a channel through which sensitive technology leaves the country.

While discussions remain ongoing, officials have not reached a final decision on the scope or timing of any new regulations. Sources indicated that any restrictions may primarily apply to future frontier AI models rather than systems already publicly available.

The discussions come as Chinese AI companies rapidly expand their global influence following the emergence of DeepSeek’s R1 model last year.

Chinese developers have gained international attention by producing capable large language models at significantly lower costs than many leading American systems.

Among the companies participating in the government consultations are some of China’s most influential AI developers. Alibaba’s Qwen family of models has become one of the country’s most widely adopted AI platforms, while ByteDance has aggressively expanded its Doubao model across consumer and enterprise applications.

Meanwhile, startup Z.ai has attracted growing attention from international AI researchers after its GLM-5.2 model demonstrated performance approaching some of Silicon Valley’s leading systems while requiring substantially fewer computing resources.

These advances have strengthened China’s position in the global AI race and increased Beijing’s incentive to protect technologies viewed as strategically valuable.

Potential Global Implications

Any restrictions on international access to Chinese frontier AI models are expected to have significant consequences for businesses worldwide. Many companies have incorporated Chinese models into their AI strategies because they often deliver competitive performance at considerably lower operating costs.

Industry experts note that limiting overseas availability would reduce competition in the global AI market, potentially increasing costs for businesses that have benefited from inexpensive Chinese alternatives. It could also accelerate the growing technological divide between competing AI ecosystems centered around the United States and China.

China’s deliberations mirror similar measures already introduced by the United States. The administration of President Donald Trump has increasingly framed advanced AI as a national security issue, particularly over concerns that frontier American models could be exploited by foreign militaries or intelligence agencies.

In June, Washington imposed export restrictions on Anthropic’s advanced Fable and Mythos models, initially prompting the company to disable access globally because it could not reliably verify users’ nationalities in real time. While restrictions on the general-purpose Fable model were later eased after additional safeguards were introduced, access to the cybersecurity-focused Mythos model remains limited to selected trusted U.S. organizations.

American policymakers have also debated whether additional safeguards are needed to limit the use of Chinese-developed AI systems inside the United States.

According to two of the sources, Chinese officials are particularly concerned about the cybersecurity capabilities of Anthropic’s Mythos model. Authorities reportedly fear the model could identify software vulnerabilities that might eventually be exploited against Chinese government agencies, critical infrastructure or strategic industries.

Those concerns have been echoed publicly by Chinese cybersecurity experts, including Zhou Hongyi, founder of cybersecurity company 360, who has argued that China must develop domestic AI systems capable of matching advanced American cybersecurity models.

Broader Regulatory Tightening

The latest discussions form part of a broader campaign by Beijing to tighten control over sensitive technologies. Earlier this year, Chinese regulators reportedly ordered Meta to unwind its proposed $2 billion acquisition of Chinese-founded AI startup Manus over national security concerns.

Authorities have also introduced broader regulations governing overseas transactions involving Chinese technology, data, and strategic investments.

According to multiple sources, investigators have examined whether Manus and several other Chinese AI startups that expanded internationally may have violated export control regulations.

Although officials have not disclosed how any future restrictions on AI exports would operate, legal experts advising Chinese policymakers have proposed a tiered regulatory framework.

According to discussions published in an official journal of China’s Supreme People’s Court, basic open-source AI models could remain widely available after simple regulatory filings, while more capable systems would require national security reviews. Under the proposal, the most advanced frontier AI models could be prohibited from public release altogether or made available exclusively within China.

If adopted, such a framework would represent one of the world’s most comprehensive attempts to regulate cross-border access to advanced artificial intelligence, further bolstering the technological rivalry between the world’s two largest economies.