Beijing: Chinese military linked researchers have reportedly been using advanced artificial intelligence models developed in the United States to help build and improve AI systems for defence related purposes, according to a Reuters investigation. The findings have raised fresh concerns in Washington about how cutting edge American technology may be helping China's military progress despite strict export controls and growing restrictions on access to advanced computer chips.
The Reuters investigation examined more than 80 research papers, patent applications and technical documents published by researchers associated with Chinese military institutions. The report found that many of these studies described the use of a technique known as model distillation, which allows developers to create smaller AI models by learning from the responses produced by much larger and more advanced systems. Rather than copying the original software or source code, the smaller models are trained using the answers generated by leading AI systems.
Researchers linked to the People's Liberation Army, the National University of Defense Technology and other military affiliated organisations reportedly explored using these distilled AI models for a range of defence applications. These include battlefield intelligence, surveillance, command support, cyber operations and the development of autonomous systems such as drones. Experts say that while distilled models may not match the full capabilities of the original systems, they can still perform highly specialised tasks effectively while requiring far less computing power.
The report comes as the United States and China continue to compete for leadership in artificial intelligence. Washington has introduced several measures over the past few years to limit China's access to advanced semiconductor chips and high end AI technologies. These restrictions were designed to slow China's progress in developing powerful AI systems that could have military applications. However, the Reuters findings suggest that Chinese researchers may have found alternative ways to improve domestic AI capabilities by making use of publicly available AI services.
Model distillation itself is a widely accepted machine learning technique used by universities and technology companies around the world. The growing concern is not about the method itself, but whether proprietary AI models owned by American companies are being used without permission to develop competing technologies or military systems. Legal experts say the issue could lead to further debate over intellectual property rights, licensing agreements and the terms of service governing access to commercial AI platforms.
The investigation also comes at a time when the United States is increasing scrutiny of Chinese AI companies. Earlier this week, officials in the Trump administration signalled that they were considering investigations into allegations that some Chinese firms may have used American AI models to train their own systems without proper authorisation. Chinese AI company Moonshot AI has denied claims that its recently released Kimi K3 model was built using unauthorised distillation of technology developed by American AI company Anthropic.
China has rejected accusations that it is unfairly benefiting from American AI technology. Chinese officials have accused the United States of trying to maintain technological dominance through export controls and sanctions. Beijing has warned that it will respond if additional restrictions are imposed on its AI industry.
The latest findings also follow earlier reports that Chinese military researchers had adapted open source AI models for defence purposes. Previous investigations showed that researchers connected to China's military had modified publicly available language models despite licensing restrictions intended to prevent military use. Those reports had already raised questions about how effectively AI companies can enforce restrictions once their technology becomes widely available.
Artificial intelligence is increasingly becoming a central part of national security strategies around the world. Both the United States and China are investing heavily in AI for military planning, intelligence analysis, logistics and autonomous systems. Governments believe AI will play a major role in future defence capabilities, making competition in this field one of the most important aspects of the wider technological rivalry between the two countries.
While the investigation does not suggest that Chinese researchers obtained the confidential source code or internal architecture of leading American AI models, it highlights how publicly accessible AI systems can still contribute to the development of powerful domestic technologies. The findings are expected to fuel further debate over AI security, international regulations and the challenges of preventing advanced technologies from being adapted for military use.