China appears to be rapidly expanding its military use of artificial intelligence (AI) around its homegrown DeepSeek model.
A Reuters report suggests that Beijing is heavily adopting DeepSeek to close the technology gap with the US and pursue what it calls “algorithmic sovereignty.”
The aim is to curb reliance on Western tech and strengthen domestic control over critical digital infrastructure, all while advancing capabilities in autonomous target recognition, real-time battlefield decision-making, and AI-assisted planning across the People’s Liberation Army (PLA).
Amid the growing pursuit of AI in military use, Chinese officials have committed that humans will retain control over weapons systems.
DeepSeek in Defense
DeepSeek’s growing role within the PLA has surfaced through new military assets, procurement tenders, and research projects.
In February, state-owned Norinco unveiled the P60 military vehicle, capable of autonomous combat-support operations at speeds of up to 50 kilometers (31 miles) per hour and powered by China’s flagship AI model.
Government officials hailed it as proof of the country’s progress in AI-enabled defense modernization.
Three months later, Beijing reportedly began using DeepSeek to support the design of a new fighter jet.
Reuters found a dozen PLA tenders filed in 2025 referencing DeepSeek, compared to only one referencing Alibaba’s rival model, Qwen, indicating the model’s active use and growing preference in military systems.
Meanwhile, researchers at Xi’an Technological University said their DeepSeek-powered system could analyze 10,000 battlefield scenarios in 48 seconds, cutting down a process that would otherwise take human planners roughly two days.
In addition, a 2025 patent filing from Beihang University detailed how DeepSeek was used to enhance drone swarm coordination and target tracking for small, slow, and low-flying aircraft.
“DeepSeek has willingly provided, and will likely continue to provide, support to China’s military and intelligence operations,” a US State Department spokesperson told Reuters.
Hardware Tug-of-War
Despite US export restrictions, the PLA and its research affiliates reportedly continue to use US-made Nvidia A100 and H100 chips for AI model training and experimentation.
It was not clear if the chips were stockpiled before the ban or obtained through secondary channels afterward.
Reuters identified 35 military-linked patent filings citing Nvidia A100 use and another 15 filings referencing Huawei’s Ascend chips as domestic alternatives.
The PLA Rocket Force University of Engineering also filed a 2025 patent for a remote-sensing target detection system trained on Nvidia hardware.
Meanwhile, Chinese firms such as Shanxi 100 Trust Information Technology and Landship Information Technology have claimed to use Huawei chips for AI integration in vehicles and imagery analysis systems.
Neither the Chinese nor US governments have commented on the issue, but Nvidia spokesman John Rizzo told Reuters that reselling old chips “doesn’t enable anything new or raise national security concerns,” adding that using restricted products for military use “would be a nonstarter without support or maintenance.”
