Chinese military researchers use US AI to train defense systems - Reuters
Photo: China denies conducting such practices (Getty Images)
Chinese military researchers are using the results of work from leading American AI models, including OpenAI and Anthropic, to train their own AI for the purpose of enhancing China's defense capabilities, according to Reuters.
As stated in the article, the agency reviewed over 80 Chinese scientific papers and patents that show how military and security-related institutions in China are using advanced US artificial intelligence to rapidly develop their own specialized systems.
The reviewed documents demonstrate widespread use of the "model distillation" method, in which the output of a powerful AI system is used to train smaller models that can be deployed locally without incurring the massive computing costs of building advanced AI systems from scratch.
The materials state that China's defense agencies view advanced American AI models as a source of technical knowledge and a way to narrow the gap with American competitors.
For example, in a 2025 article by researchers from Unit 96941 of China's People's Liberation Army, which handles military intelligence and cyber warfare in Beijing, the use of OpenAI's GPT-3.5 to process classified military source material was described.
The researchers stated that third-party models are unsuitable for processing classified information. To overcome this limitation, they used GPT-3.5 to compile brief descriptions of software code.
Thus, using these descriptions, they trained a domestic AI model intended to operate exclusively within Chinese military networks.
Other examples
According to analyses of scientific papers, Chinese researchers have used distillation for various purposes, ranging from content monitoring to military force deployment.
For example, at North China University, which is closely tied to the country's arms industry, researchers used the Claude 3 Haiku library from Anthropic to generate synthetic training data for a text-classification model for social media monitoring and content moderation.
Another example is a 2024 article published by the National University of Defense Technology of the People's Liberation Army of China. This material describes the use of distillation to reduce the size of an image-processing model, which in turn enables drones to analyze video in real time and support navigation and targeting decisions even when communication links are lost.
Similarly, as mentioned in a 2026 study by the Chinese Academy of Military Sciences, researchers used distillation to deploy a target recognition model on tactical equipment during simulated operations involving drones, ships, and unmanned underwater vehicles.
What the US and China are saying
According to Reuters, the issue has become one of the main sources of tension ahead of the US-China talks on AI governance and security scheduled for September.
US officials have accused some Chinese companies of using a technique known as distillation to replicate the capabilities of American AI models. Washington argues that the practice could undermine US export controls and violate intellectual property rights.
China has rejected the allegations, accusing the US of pursuing AI hegemony. Beijing also claims that American companies have engaged in similar practices.
What else you should know
As reported yesterday, Chinese President Xi Jinping has ordered the military to accelerate the adoption of drones and artificial intelligence as part of Beijing's push to build a more advanced armed forces.
Reuters also reported that the US is preparing to ban imports of new Chinese robots and power inverters. The move is intended to protect the development of American AI from national security risks while helping bring back key strategic industries expected to drive future growth.
