China uses US AI for defence systems
Chinese military researchers have tapped outputs from leading U.S. artificial intelligence models to train domestic systems, according to a review of more than 80 Chinese academic papers and patents.
U.S. models used as a shortcut
The analysis, compiled from documents supplied by a Washington‑based group, shows that teams linked to the People’s Liberation Army are employing a technique called “model distillation.” In this process, a powerful AI system’s outputs are fed into smaller, specialized models that can run on local hardware without the massive computing power required to build frontier AI from scratch.
One paper from a PLA cyber‑warfare unit described using OpenAI’s GPT‑3.5 to summarize sensitive military source code. The summaries then trained a domestic model that operates entirely within Chinese military networks, avoiding reliance on third‑party services for classified work.
At the North University of China, scholars used Anthropic’s Claude 3 Haiku to generate synthetic data for a text‑classification model aimed at social‑media monitoring and content moderation. Anthropic has said it does not grant commercial access to Claude in China, and it monitors for policy violations, raising questions about whether distilled models retain original safety safeguards.
Applications across the defence spectrum
Distillation appears in projects ranging from content monitoring to frontline deployments. Researchers at the National University of Defense Technology reported shrinking an image‑processing model for use on unmanned aerial vehicles, enabling drones to analyse live video and support navigation and targeting even when communications are cut.
Related: Rhitik Jassar Launches ZTA Infrastructure
Similarly, the Academy of Military Sciences detailed a target‑recognition model run on tactical hardware during simulated maritime operations involving drones, ships and unmanned submarines. These efforts suggest a systematic push to embed AI capabilities into platforms that cannot host large‑scale models.
Chinese officials have framed the practice as part of a broader push for “model lightweighting” and edge computing. Subsidies and research funding are being directed toward technologies that let AI run on devices with limited processing power, a response to U.S. export controls on high‑end chips.
The papers often note that “teaching a model the right answer is one thing but teaching it the reasoning behind the answer is much harder,” highlighting the technical challenge of capturing underlying logic.
From a practical standpoint, this approach could mean that soldiers on the ground gain access to AI‑driven tools that were previously only available to well‑funded research labs. By embedding distilled models into field equipment, the Chinese military may achieve faster decision cycles without depending on external cloud services.
Legal and security concerns
U.S. officials have accused some Chinese entities of using distillation to extract capabilities from American AI models, potentially breaching export controls and intellectual property rights. Beijing has rejected those claims, describing the accusations as part of U.S. “AI hegemonism.”
Related: How to have fun choosing your lab grown diamond engagement rings
In January, researchers at the Army Engineering University published a paper on “data‑free distillation,” a method of reverse‑engineering a model’s abilities without direct access to its parameters. They also proposed defenses to mask hidden logical information that might be exposed through public outputs.
Experts caution that distilled models inherit only selected capabilities and cannot fully replicate the breadth of frontier systems. Trevor Koverko, co‑founder of an AI data firm, said such models are “less capable than their teacher models” and should be seen as a way to transfer specific functions into cheaper, locally controlled systems.
Anthropic noted that distilled models might lose original safety features, potentially allowing sensitive capabilities to be transferred to models beyond its control. This raises concerns about the unintended spread of advanced AI functions.
Overall, the findings provide a rare glimpse into how Chinese defence institutions are leveraging U.S. AI advancements as a shortcut to narrow the technological gap, despite ongoing geopolitical tensions over AI governance and export restrictions.
