China Uses US AI Model Outputs to Train Domestic Rivals as Chip Restrictions Tighten
Chinese researchers and technology companies are increasingly using outputs from leading American artificial intelligence models to improve homegrown AI systems. The approach, known as AI distillation, is gaining traction as China faces tighter restrictions on access to advanced computing chips.
A Reuters review of more than 300 Chinese academic papers, patents, corporate disclosures, recruitment notices and government documents found that the practice has expanded beyond academic research into industrial applications.
The findings highlight how Chinese developers are exploring alternative ways to build AI capabilities while reducing their dependence on expensive computing resources.
What Is AI Distillation?
AI distillation is a training technique in which a powerful model, known as the teacher, generates answers, code or demonstrations that are then used to train a smaller student model.
Instead of repeatedly using an expensive advanced model, developers can retain selected outputs as training examples. This allows smaller systems to learn specific capabilities at a lower operating cost.
However, distillation does not reproduce the complete capabilities of a leading AI model or remove the need for advanced computing infrastructure.
OpenAI Models Used in Chinese AI Research
The Reuters investigation identified several research projects involving outputs from American AI models.
A November 2024 paper from Shanghai Jiao Tong University researchers described selecting around 5,000 answers generated by OpenAI's o1 model. The researchers translated the material using GPT-4o mini and assembled 10,750 bilingual training examples to further train Alibaba's Qwen2.5-72B model.
In another project published in April 2025, Alibaba researchers described using GPT-4 and GPT-4o to generate multilingual training material before refining and scoring the responses for smaller Qwen models.
A February 2026 research paper involving Shanghai Jiao Tong University and Meituan also described using GPT-4o-generated plans to help train a smaller model to coordinate AI agents working on programming tasks.
These examples demonstrate how outputs from advanced models can become reusable training material for domestic AI development.
Chip Restrictions Add Pressure to China's AI Strategy
US restrictions on advanced semiconductor exports have made access to high-performance computing a major challenge for Chinese AI developers.
Distillation offers one way to preserve selected capabilities while operating smaller and less expensive models. China is also encouraging the development and deployment of AI across manufacturing and other industries.
A national AI-plus-manufacturing policy package published by eight Chinese government agencies in January supports model training, fine-tuning and distillation for industrial applications.
Nevertheless, the technique cannot eliminate China's need for advanced chips or replace independent research and innovation.
Growing Tensions Over AI Intellectual Property
The growing use of American model outputs has become a point of disagreement between Washington and Beijing.
US officials have raised concerns about industrial-scale collection of outputs from leading AI systems, warning that certain activities could cross into intellectual-property theft. Chinese officials have defended distillation as a standard AI development technique.
The distinction is important: distillation itself is not inherently illegal, but the methods used to obtain and reuse model outputs may raise contractual or intellectual-property concerns.
OpenAI and Anthropic have restrictions in their terms concerning the use of model outputs to develop competing systems, subject to their respective terms and exceptions.
What This Means for the Global AI Industry
The growing adoption of distillation highlights a shift in how countries and companies approach AI development. Access to advanced chips remains important, but training data, model efficiency and software engineering are also becoming critical competitive factors.
For the global AI industry, the issue raises questions about model ownership, intellectual property, competition and the future effectiveness of technology export restrictions.
China's approach could help accelerate the adoption of lower-cost AI systems across industries. At the same time, researchers have warned that excessive reliance on foreign model outputs could limit original innovation and leave domestic developers dependent on breakthroughs made elsewhere.



