Chinese military employs model distillation of US AI

Chinese military researchers have used outputs from leading U.S. artificial intelligence models, including those developed by OpenAI and Anthropic, to train domestic AI systems aimed at advancing the country's defense capabilities. The findings, based on a review of more than 80 Chinese academic papers and patents, reveal a widespread reliance on a technique known as model distillation to circumvent restrictions on advanced computing hardware.

The technique involves using the outputs from powerful, frontier AI systems to train smaller, specialized models that can be deployed locally without the enormous computational resources required to build such systems from scratch. The reviewed documents indicate this method is widely used by researchers linked to the People's Liberation Army and other military institutions. The papers suggest Chinese defense institutions view leading U.S. AI models as both a source of technical insight and a strategic tool to close the technological gap with American rivals.

The practice has become a major point of contention ahead of U.S.-China talks on AI governance and safety. U.S. officials have accused some Chinese entities of using distillation to...