Chinese military researchers use US AI models for defense development
Researchers reportedly analyzed over 80 Chinese military research papers and patents showing that Chinese military institutions systematically extracted outputs from US AI models developed by OpenAI and Anthropic, using model distillation to train specialized domestic defense systems. The practice raises concerns about intellectual property violations and circumvention of US export controls, though China disputes the allegations and Chinese AI developers have denied their systems rely on foreign models.
What happened
Analysis of over 80 Chinese research papers and patents revealed that Chinese military researchers used outputs from leading American AI models developed by OpenAI and Anthropic to train domestic AI systems for enhancing China's defense capabilities.
The analysis showed widespread use of model distillation technique, where outputs from a powerful AI system are used to train smaller, specialized models that can be deployed locally without requiring the vast computing resources necessary to build advanced AI systems from scratch.
Reported impact
The practice reportedly represents a strategy to close the technology gap between Chinese and American competitors, with Chinese defense institutions viewing advanced American AI models as a source of technical knowledge.
Index analysis
Researchers affiliated with the People's Liberation Army and other military institutions systematically monitored reasoning steps in Western models and adapted them for surveillance, cyber warfare, and tactical decision-making purposes.
American officials accused some Chinese entities of using distillation to extract capabilities from US AI models, which could undermine export controls and violate intellectual property rights. China rejected these allegations, arguing that the US seeks dominance in AI while claiming American companies have engaged in similar practices.
Governance controls
The disagreement centers on unauthorized extraction rather than distillation itself, which is a common practice in the sector. This issue emerged as a key point of contention ahead of US-China discussions on AI governance and security.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.