Chinese military researchers have leveraged publicly available AI models from OpenAI and Anthropic to train specialized defense systems through a technique called model distillation, according to a review of over 80 Chinese academic papers and patents.
This allows China to develop advanced military AI capabilities without building frontier systems from scratch, effectively circumventing U.S. efforts to restrict Beijing's access to advanced chips and strategic technologies.
What happened
Chinese military researchers have used outputs from OpenAI and Anthropic AI models to train domestic defense systems, according to a review of more than 80 Chinese academic papers and patents. The researchers employed "model distillation"—a technique that takes outputs from powerful AI systems to train smaller, specialized models that can run locally without massive computing power.
Why it matters
China's military and security institutions are using U.S. AI as a shortcut to develop their own advanced systems, sidestepping Washington's restrictions on Beijing's access to cutting-edge chips and strategic technologies. This reveals how trade controls may have unintended gaps when companies release their models publicly.
What to watch
The practice is widespread among researchers linked to the People's Liberation Army and other Chinese military institutions, suggesting this is not isolated. The Jamestown Foundation's compilation of these papers and patents indicates the scale is significant enough to warrant scrutiny from U.S. policymakers.
Chinese military researchers have found a way to leverage cutting-edge U.S. artificial intelligence models as a shortcut to building their own advanced defense systems. According to a review of more than 80 Chinese academic papers and patents conducted with research compiled by the Washington-based Jamestown Foundation, military and security-linked institutions in China are systematically using outputs from OpenAI and Anthropic AI models to train domestic systems. This practice represents a significant challenge to U.S. export control efforts, which have focused on restricting Beijing's access to advanced semiconductor chips and other strategic technologies.
The technique underlying this activity is known as "model distillation." In this process, the outputs from a powerful, frontier AI system are used to train smaller, specialized models that can be deployed and run locally without the enormous computing requirements needed to build such systems from scratch. Model distillation acts as a knowledge-transfer mechanism, allowing researchers to extract and compress the capabilities of large models into more compact forms suitable for military applications. The review of academic papers and patents shows that this technique is not limited to a handful of researchers but is widely used across institutions linked to the People's Liberation Army and other Chinese military bodies, indicating a coordinated and systematic approach to leveraging U.S. AI technology for defense advancement.
The findings underscore a significant gap in the effectiveness of U.S. export controls aimed at restricting China's access to advanced AI technology. While Washington has focused on limiting Beijing's ability to acquire cutting-edge semiconductors and frontier AI systems, the review reveals that publicly released AI models from leading U.S. companies—OpenAI and Anthropic—are being repurposed by China's military apparatus to accelerate its own capabilities. Model distillation, a well-known AI technique, allows researchers to extract the knowledge encoded in large, resource-intensive models and compress it into smaller systems that can run on standard hardware. This approach dramatically lowers the barrier to entry for developing specialized military AI applications.
The scale of this activity—evidenced by the compilation of more than 80 academic papers and patents—suggests this is not an isolated experiment but a systematic effort. The involvement of institutions directly linked to China's military establishment indicates that these tools are being weaponized for defense purposes, not merely academic exploration. The body's reference to the Washington-based Jamestown Foundation's compilation further suggests that U.S. intelligence and policy communities are aware of the trend and view it as significant enough to merit public attention.
AI-summarized, only the topics you pick — one digest a day via Email, Slack, or Discord.
Free · takes 30 seconds · unsubscribe anytime
Ask AI anything about this article. Q&As are published on this page for other readers too.
The AI news that matters, in one minute each morning.
Sign up free