# China Considers New Controls on Exports of Frontier AI Models > Source: > Published: 2026-07-07 19:44:34+00:00 China is weighing a significant shift in AI policy that could limit overseas access to its most advanced AI models, suggesting that Beijing views frontier AI as a strategic technology requiring tighter government control. The discussions, first reported by Reuters, could affect both proprietary models and open-weight systems that have supported China’s expansion in the global AI market. Officials from China’s Ministry of Commerce have met over the past month with executives from Alibaba, ByteDance and AI startup Z.ai to discuss possible restrictions on foreign access to current and future AI models. While no final decisions have been made, the talks indicate concern over protecting domestic AI capabilities as competition with the US intensifies. The proposals under discussion extend beyond limiting exports, and also examined whether theft or unauthorized disclosure of proprietary AI technology should fall under China’s national security law, meaning offenders could face much tougher penalties. Another proposal would impose additional scrutiny over who can finance Chinese AI startups, adding still more government oversight to an industry already subject to expanding regulation. The discussions reportedly covered both closed-source systems and open-weight models that developers can download, customize and deploy themselves. That distinction is important because Chinese companies have built an international following by releasing capable models at relatively low cost. Alibaba’s Qwen family, ByteDance’s Doubao platform and Z.ai’s GLM-5.2 have attracted developers looking for alternatives to higher-priced US offerings. China’s emerging AI security approach parallels measures already adopted in Washington. The US has expanded export controls on advanced AI technologies and recently restricted access to some of Anthropic’s most advanced models over national security concerns. **A Rising AI Sector ** China’s AI industry has gained momentum since DeepSeek demonstrated that competitive LLMs could be built at substantially lower cost than many Western rivals. Affordable pricing and improving technical performance have encouraged businesses worldwide to adopt Chinese models, particularly for enterprise development projects. Any restrictions on overseas availability could therefore boost prices for companies that have integrated these models into their AI strategies. Although authorities have not disclosed how any future restrictions would operate, one possible framework has already surfaced. Legal scholars participating in a May policy discussion proposed categorizing AI systems according to their sensitivity. Under that approach, lower-risk open-source software would require only basic registration, more advanced models would undergo security reviews, and the most capable frontier systems could remain available only within China or be blocked from public release altogether. The policy discussions are similar to an effort by China to tighten oversight of emerging AI technologies. Earlier this year regulators investigated several domestic AI startups that relocated operations overseas to determine whether export control rules had been violated. Authorities also strengthened reviews of international transactions involving Chinese technology and sensitive data, while reportedly requiring some AI startups to obtain approval before accepting US investment. Chinese officials have reportedly expressed concern that advanced US AI systems, including Anthropic’s cybersecurity-focused Mythos model, could be used to identify software vulnerabilities that might be exploited against Chinese infrastructure. Those concerns have been echoed publicly by Chinese cybersecurity executives and state media, supporting the government’s emphasis on developing domestic alternatives in strategically important AI capabilities.