Back
China considers curbing overseas access to its most advanced AI models

China considers curbing overseas access to its most advanced AI models

As Reuters reports talks with Alibaba, ByteDance and Z.ai on tiered AI export controls, cheaper Chinese models from firms such as DeepSeek and Moonshot AI are gaining U.S. business users.

Fact Check
The Reuters exclusive dated July 7, 2026 directly confirms that Beijing's Ministry of Commerce held talks with Alibaba, ByteDance and Z.ai about curbing overseas access to China's most advanced AI models, including discussion of a tiered regulatory approach for open-source models. The WSJ article corroborates both the policy consideration and the detail that cheaper Chinese models from DeepSeek and Moonshot AI are gaining US business users. The CryptoBriefing lead restates the same reporting. All components of the claim are supported by these sources; the only minor caveat is that the policy is still under discussion rather than finalized, which the claim accurately reflects ('considers').
Summary

China has been discussing possible restrictions on overseas access to its most advanced AI models in meetings led by the Ministry of Commerce with companies including Alibaba, ByteDance and Z.ai, according to Reuters. The talks have covered both closed-source and open-weight models, including unreleased systems, and participants discussed a tiered framework under which basic open-source tools would require filing, more advanced technologies would undergo security reviews, and the most sensitive frontier models could be barred from public release or limited to domestic use. Reuters also reported that officials raised making unauthorized disclosure or theft of proprietary AI technology an offense under China’s national security law and discussed possible limits on which investors can fund domestic AI startups. No timeline has been set and it is unclear whether any measures will take effect; two sources told Reuters said the rules may apply only to future models rather than those already released. The potential curbs come as U.S. companies increasingly use lower-cost Chinese AI models, including products from DeepSeek and Moonshot AI, alongside offerings from OpenAI and Anthropic.

Terms & Concepts
  • open-weight models: AI models whose trained parameters are made available so others can run, adapt or build on them.
  • frontier models: The most advanced AI systems at the leading edge of capability and performance.
  • security reviews: Government assessments used to evaluate whether a technology can be released without creating national security risks.