China uses US AI models to train domestic rivals as technology tensions grow

China uses US AI models to train domestic rivals as technology tensions grow

Beijing: Chinese researchers and technology companies have been using advanced artificial intelligence models developed by American firms to train and improve their own systems, according to an investigation. The investigation reviewed more than 300 documents, including academic papers, patents, company disclosures, recruitment notices and government records. It found that the practice, known as knowledge distillation, is increasingly being used beyond academic research and in commercial AI development.

Knowledge distillation involves using a more advanced AI model to generate answers, code, explanations or other information that can be used to train a smaller model. This allows developers to transfer some capabilities from an established system to a less powerful one without repeating the entire process of developing a model from the beginning. One example involved researchers at Shanghai Jiao Tong University. In November 2024, they selected about 5,000 responses generated by OpenAI's o1 model and translated them using GPT-4o mini. They then created 10,750 bilingual training examples to further train Alibaba's Qwen2.5-72B model.

In April 2025, Alibaba researchers also used GPT-4 and GPT-4o to generate multilingual training material for smaller Qwen models. Another project, conducted in February 2026 by researchers from Shanghai Jiao Tong University and Chinese food delivery company Meituan, used plans generated by GPT-4o to train a smaller model to coordinate multiple AI agents working on programming tasks. The growing use of American AI models has raised concerns in Washington about the transfer of technology and the protection of intellectual property.

In September 2026, the US Cybersecurity and Infrastructure Security Agency, the National Security Agency and the Federal Bureau of Investigation accused Chinese companies of carrying out targeted, large scale efforts to extract capabilities from American AI systems. The agencies named DeepSeek, Moonshot AI, Alibaba, MiniMax, StepFun and Z.AI. They alleged that billions of tokens had been exchanged across millions of interactions since at least late 2024. These allegations have not been established as judicial findings. China’s Ministry of Commerce rejected the accusations, describing them as groundless and legally unsound. It defended knowledge distillation as a normal technology practice and accused the United States of politicising technological competition.

American AI companies have also introduced terms restricting the use of their services and outputs to develop competing models. However, a breach of a company's terms of service does not automatically amount to criminal intellectual property theft. The dispute is unfolding alongside wider US efforts to restrict China's access to advanced computing technology. On November 15, the US Department of Commerce's Bureau of Industry and Security announced new rules addressing remote access to advanced American semiconductors and computing systems through overseas data centres. The measures require data centres serving companies in China, Russia, Iran and Cuba to identify their end users.

Additional case by case licensing requirements also apply to certain chip sales involving Indonesia, Malaysia, Singapore, Thailand and Brazil. The restrictions are intended to prevent advanced American computing technology from reaching restricted users through third countries. Chinese Foreign Ministry spokesperson Guo Jiakun criticised the measures, describing them as economic coercion and excessive extraterritorial jurisdiction. He said China would take necessary measures in response.

China is also seeking to expand the use of artificial intelligence across its domestic industries. In January 2026, the country published an action plan targeting the development of three to five general purpose AI models for manufacturing by 2027. It also aims to establish 100 high quality datasets, develop 500 typical application scenarios and support two to three internationally influential companies in the sector. The developments highlight the growing competition between the world's two largest economies over artificial intelligence, advanced semiconductors and technological independence.

The issue comes as leaders from the Asia Pacific Economic Cooperation forum prepare to meet in Shenzhen, China, on November 18 and 19. Chinese President Xi Jinping is chairing the meeting, with US President Donald Trump expected to meet him during the visit. The discussions are taking place amid wider tensions over trade, technology and access to advanced computing resources.


Follow the CNewsLive English Readers channel on WhatsApp:
https://whatsapp.com/channel/0029Vaz4fX77oQhU1lSymM1w

The comments posted here are not from Cnews Live. Kindly refrain from using derogatory, personal, or obscene words in your comments.