Chinese Military Researchers Use US AI Models to Develop Defence Systems

Members of China's People's Liberation Army participate in a military event as researchers continue developing domestic AI systems using model distillation techniques.

Chinese Military Researchers Use US AI Models to Advance Defence AI

BEIJING — July 31, 2026

Chinese military researchers have used leading U.S. artificial intelligence (AI) models to help train domestic AI systems for defence-related projects, according to a review of more than 80 Chinese academic papers and patents.

The documents indicate that researchers linked to China's military have applied a technique known as model distillation, which uses the outputs of powerful AI models to train smaller, specialized systems. These compact models require far less computing power and can operate within secure local networks.

The findings suggest Chinese defence institutions are using advanced U.S.-developed AI systems as a foundation for developing tools tailored to military operations, despite U.S. efforts to restrict China's access to advanced semiconductor technology and other strategic resources.

The reviewed papers describe AI applications across several defence areas, including surveillance, cybersecurity, software analysis, drone operations and battlefield decision-making. Researchers appear to use the distilled models to improve efficiency while keeping sensitive work inside domestic military systems.

One study published last year described how researchers from PLA Unit 96941 used OpenAI's GPT-3.5 to analyze military software code. Instead of processing classified information directly through an external AI service, they generated summaries using the model and trained a domestic AI system to perform similar tasks entirely within Chinese military networks.

Other research showed that scientists at the National University of Defense Technology developed a lightweight image-processing model for unmanned aerial vehicles. The system was designed to analyze live video and support navigation and targeting even when communications are disrupted.

Another study from the Academy of Military Sciences described using a distilled AI model for target recognition during simulated maritime operations involving drones, ships and unmanned submarines.

The documents also show that military-linked institutions are not the only organizations using the technique. Researchers at the North University of China employed Anthropic's Claude 3 Haiku model to generate synthetic training data for systems used in social media monitoring and content moderation.

Anthropic said it does not provide commercial access to its AI models in China or to organizations controlled by Beijing and uses monitoring systems to detect possible policy violations. The company also warned that distilled AI models may lose important built-in safety protections during the training process.

The issue has become a growing point of tension between the United States and China as competition over artificial intelligence intensifies. U.S. officials have accused some Chinese organizations of using model distillation to reproduce capabilities from American AI systems in ways that may undermine export controls and intellectual property protections.

China has rejected those accusations, arguing that the United States is attempting to dominate global AI development while claiming that American companies have also benefited from similar industry practices.

Experts note that model distillation can transfer selected capabilities from advanced AI systems but cannot fully reproduce the broad performance of the original models. Building world-leading AI systems still requires enormous computing resources, advanced semiconductor chips and extensive research.

Researchers in China have also begun studying the security risks of model distillation itself. A paper published earlier this year proposed methods to reduce the amount of internal reasoning that could be extracted from publicly available AI models, reflecting concerns that the same technique could also be used against Chinese AI systems.

The findings highlight how artificial intelligence has become an increasingly important area of strategic competition, with governments seeking both technological leadership and stronger safeguards for advanced AI systems.