China taps US AI models to train homegrown rivals as chip curbs bite
QQQ•Chinese developers are increasingly using outputs from leading US AI models to train domestic rivals, as US chip restrictions constrain access to advanced computing. A review of more than 300 documents shows the practice has spread into industrial use and is encouraged in policies supporting AI deployment.
1. Distillation expands in China
Chinese developers are using leading US AI models to generate answers, code and solutions that can train smaller domestic systems. A review of more than 300 academic papers, patents, corporate disclosures, recruitment notices and government documents found the practice has moved beyond academic experiments into mainstream industrial use.
2. Examples in research
A November 2024 paper described Shanghai Jiao Tong University researchers selecting about 5,000 answers from OpenAI’s o1, translating them with GPT-4o mini and assembling 10,750 bilingual training examples for Alibaba’s Qwen2.5-72B model. An April 2025 paper described Alibaba researchers using GPT-4 and GPT-4o to generate and refine training material for smaller Qwen models. Researchers from Shanghai Jiao Tong University and Meituan also described using GPT-4o to help train a smaller model to coordinate AI agents on programming problems.
3. Policy and US concerns
Chinese government policies encourage distillation for industrial applications and wider AI deployment, while US officials have warned that industrial-scale collection of model outputs could amount to intellectual-property theft. Chinese officials defended distillation as a standard development technique, and Chinese developers have sought guidance from Beijing on the boundaries of ordinary use and organised collection. Distillation is not illegal, but it cannot reproduce a leading model’s full capabilities or eliminate the need for advanced computing power.




