Chinese military AI distillation is at the center of new findings that researchers in China’s defense ecosystem have used outputs from leading U.S. artificial intelligence models, including those from OpenAI and Anthropic, to train domestic systems intended to bolster the country’s military capabilities, according to a review of more than 80 Chinese academic papers and patents.
The review offers a rare look at how security-linked institutions in China are tapping state-of-the-art U.S. AI as a shortcut to develop specialized tools at home, even as Washington seeks to limit Beijing’s access to advanced chips and other strategic technologies.
Chinese military AI distillation detailed across papers
The documents describe extensive use of “model distillation,” a technique in which outputs from a powerful AI system are used to train smaller, task-specific models that can run locally without the heavy computing needed to build frontier systems. Research compiled by a Washington-based think tank and verified by reporters indicates distillation is widely employed by researchers connected to the People’s Liberation Army and other military bodies.
The papers indicate Chinese defense institutions view top U.S. AI systems as both technical guides and tools to narrow the performance gap with American counterparts. The debate focuses on alleged unauthorized extraction of capabilities, not on distillation itself, which remains a common industry practice.
The issue has emerged as a flashpoint ahead of U.S.-China talks on AI governance and safety. U.S. officials have accused some Chinese entities of using distillation to pull capabilities from American models, potentially weakening export controls and infringing intellectual property rights.
Beijing has pushed back, accusing Washington of AI “hegemonism” and asserting that U.S. firms have engaged in similar conduct. Chinese developers have also rejected claims that their progress depends on foreign systems. Last week, startup Moonshot denied allegations from the Trump administration that its Kimi K3 model was built using distillation, saying it relied on proprietary innovation.
Sunny Cheung, a fellow who analyzed more than 60 of the papers, said Chinese military scientists are systematically capturing the reasoning steps of Western models to adapt them for surveillance, cyber operations and tactical decision-making. He noted that transferring the reasoning behind answers is considerably more difficult than transferring answers themselves.
Reporters verified the academic literature and identified two dozen additional military-linked case studies. One paper from 2023 by researchers in PLA Unit 96941, an intelligence and cyber-warfare unit in Beijing, described using OpenAI’s GPT-3.5 to process sensitive military source code. The team argued that third-party models could not handle classified material directly, so they used GPT-3.5 to summarize code and then trained a domestic model on those summaries to operate entirely within military networks.
The White House, the Pentagon, China’s foreign ministry, the PLA and OpenAI did not respond to requests for comment.
Applications span monitoring, drones and maritime drills
Chinese researchers have applied distillation across use cases from content monitoring to battlefield deployment, the review showed. At the North University of China, which has close ties to the country’s weapons sector, researchers used Anthropic’s Claude 3 Haiku to generate synthetic data for a text classification model used in social media monitoring and content moderation.
Anthropic said it does not provide commercial access to Claude in China or to Beijing-controlled firms, and that it uses monitoring systems to detect policy violations. The company also warned that distilled models can lose original safety guardrails, potentially transferring sensitive capabilities to systems beyond its control.
A 2024 paper from the PLA’s National University of Defense Technology detailed shrinking an image-processing model via distillation for deployment on unmanned aerial vehicles. The approach let drones analyze live video and support navigation and targeting in real time, even when communications were cut. Separately, researchers at the Academy of Military Sciences used distillation to run a target-recognition model on tactical hardware during simulated maritime operations involving drones, ships and unmanned submarines, according to a study published earlier this year.
China has embraced distillation as it competes with the United States in frontier AI while contending with limits on advanced computing imposed by U.S. export controls. Central and local authorities have promoted “model lightweighting” and edge computing, steering subsidies and research funds to technologies that allow AI to run on drones, satellites and other devices with constrained processing power.
Limits and security risks of Chinese military AI distillation
Experts caution that distillation carries significant limitations. As Chinese AI systems improve, military researchers are also scrutinizing distillation as a security concern. In January, scholars at the Army Engineering University examined the threat of “data-free distillation,” a method to reverse-engineer a model’s capabilities without direct access to its core parameters. They proposed defenses to obscure hidden logical information exposed in public outputs.
Analysts emphasize that distilled models inherit only selected capabilities and cannot fully reproduce the broad intelligence of leading systems. Trevor Koverko, co-founder of AI data firm Sapien, said distilled models remain less capable than their teacher systems, characterizing the process as a way to transfer select functions into cheaper, locally controlled models rather than to achieve independence from frontier AI.
For Hawaiʻi readers tracking U.S.-China tech competition, the findings highlight how export controls and intellectual property safeguards are being tested by fast-evolving AI practices, with potential implications for security in the Indo-Pacific. For more context on how national policy choices ripple through Hawaiʻi’s economy and security, see our coverage of Hawaii governor candidates and their positions on key energy and infrastructure issues.
