Key Takeaways:
Chinese authorities have held meetings with top tech firms over the past month about potentially restricting overseas access to the country's most advanced AI models, a move that could reshape global access to low-cost frontier AI.
Chinese authorities convened discussions with Alibaba Group Holding Ltd., ByteDance Ltd. and Z.ai Co. about limiting foreign access to China's most capable AI systems — including models not yet released — according to three people familiar with the talks who declined to be identified. The meetings, led by the Ministry of Commerce, covered both closed-source and open-weight models, the people said.
"The discussions reflect a broader recognition that frontier AI capabilities are now treated as strategic national assets requiring controls," said Scott Singer, a fellow in the technology and international affairs program at the Carnegie Endowment for International Peace. "China is going to have to balance the benefits of access to global markets with a desire to control a technology that is central for national security."
Officials raised the possibility of making any leak or theft of proprietary AI technology an offense under China's national security law, two of the people said. They also discussed new measures to restrict who can fund domestic AI startups. The scope of potential restrictions remains under discussion and may apply only to future models, the people added, with no timeline for implementation.
The talks follow a series of steps by Beijing to keep homegrown AI within the country. In April, China's state planner ordered Meta Platforms Inc. to unwind its $2 billion acquisition of Chinese-founded AI startup Manus. In early June, authorities issued sweeping new rules tightening control of overseas deals involving Chinese investors, technology and national security. China also launched investigations this year into Manus and other local AI startups that moved abroad, examining whether they violated export control laws.
The US Precedent
The potential restrictions mirror actions taken by Washington. In June, the Trump administration ordered that foreign nationals not have access to Anthropic's most advanced Fable and Mythos models, citing national security concerns about the potential for AI to be misused by military intelligence in China, Russia and other countries. Export controls for Fable, designed for the general public, were later lifted after new safeguards were put in place. But Mythos, built for cybersecurity professionals, remains available only to some "trusted" U.S. organizations.
Chinese authorities are deeply worried about Mythos' potential to exploit software vulnerabilities and that Washington might deploy the model against Chinese interests, two of the people said. Zhou Hongyi, founder of cybersecurity firm 360 Security Technology Inc., a major vendor to government and enterprise clients, has said China needs to develop its own equivalent.
Market Impact at Stake
Any decision by Beijing to limit access could ripple across global AI markets. Chinese AI models have made significant inroads since DeepSeek's R1 model emerged last year, offering capabilities approaching those of leading U.S. offerings at 60% to 90% lower cost, according to CNBC. OpenRouter, a startup that lets U.S. companies route queries across different AI models, said Chinese models now handle about 30% of its traffic.
Alibaba's Qwen and ByteDance's Doubao are among the most widely used AI models in China. Z.ai's GLM-5.2, released in late June, has drawn attention in Silicon Valley for matching Anthropic's Mythos in bug-spotting ability at a fraction of the cost. Some early-stage U.S. startups have shifted all their traffic to Chinese models, creating potential disruption if access is cut off.
A May roundtable of Chinese legal experts on open-source AI regulations, published in an official Supreme People's Court journal, proposed a tiered system: basic open-source tools subject to simple filing, more advanced technologies facing security reviews, and the most sensitive frontier models barred from public release or restricted to domestic use.
Joseph Spisak, vice president of product and head of open source at AI startup Reflection, said some firms risk becoming "locked into" the Chinese AI ecosystem if they build applications on top of those models. "Once you're hooked on the technology, there's a certain amount of inertia that gets created," he said. "It's harder and harder to move away."
This article is for informational purposes only and does not constitute investment advice.
