China mulls cutting foreign access to DeepSeek, Qwen, and other advanced AI models

Chinese authorities are weighing major restrictions on foreign access to the country's most advanced AI models, according to Reuters. Officials from the Ministry of Commerce held discussions with Alibaba, ByteDance, and Z.ai over the past month. The proposed curbs would apply to both closed and open-weight systems, covering popular models like DeepSeek, Qwen, Doubao, and GLM. One proposal would classify the leak or theft of proprietary AI technology as a national security crime; another would limit the types of investors eligible to fund domestic AI companies. A tiered system has been suggested: basic tools would need light filing, stronger models would undergo security reviews, and the most sensitive would be restricted to domestic use only. No final decisions have been made, and any restrictions may only affect future models.

These potential restrictions mark a significant shift in China's AI strategy. Previously, Chinese models have served as affordable alternatives to expensive American systems from OpenAI, Google, and Anthropic, particularly gaining traction among European developers. If implemented, the curbs could reduce supply and increase costs for businesses relying on these models. While already-downloaded models would be hard to recall, uncertainty now surrounds the global release of next-generation Chinese AI systems, mirroring US export controls on advanced chips.