China’s state-owned defense giant Norinco in February unveiled a military vehicle capable of autonomously conducting combat-support operations at 50kph. It was powered by DeepSeek, the company whose artificial intelligence (AI) model is the pride of China’s tech sector.
The Norinco P60’s release was touted by Chinese Communist Party officials in statements as an early showcase of how Beijing is using DeepSeek and AI to catch up in its arms race with the US at a time when leaders in both countries have urged their militaries to prepare for conflict.
A Reuters review of hundreds of research papers, patents and procurement records gives a snapshot of the systematic effort by Beijing to harness AI to gain a military advantage.
Specifics of how the systems behind China’s next-generation weapons work and the extent to which it has deployed them are a state secret, but procurement records and patents offer clues into Beijing’s progress toward capabilities such as autonomous target recognition and real-time battlefield decision support in a way that mirrors US efforts.
Reuters could not establish if all the products had been built and patents do not necessarily indicate operational technology.
The Chinese People’s Liberation Army (PLA) and affiliates continue to use and look for Nvidia chips, including models under US export controls, according to the papers, tenders and patents.
Reuters could not determine if those chips were stockpiled before Washington imposed restrictions, as the documents do not detail when the hardware used was exported.
Patents filed as recently as June show their use by military-linked research institutes. In September 2022, the US Department of Commerce banned exports to China of Nvidia’s popular A100 and H100 chips.
Nvidia spokesman John Rizzo said in a statement to Reuters that while the firm cannot track individual resales of previously sold products, “recycling small quantities of old, second-hand products doesn’t enable anything new or raise any national security concern. Using restricted products for military applications would be a nonstarter, without support, software or maintenance.”
The US departments of the treasury and commerce did not respond to questions about Reuters’ findings.
The Chinese military this year has also increased its use of contractors that claim to exclusively use domestically made hardware such as Huawei AI chips, said Sunny Cheung (張崑陽), a fellow at the Washington-based Jamestown Foundation defense policy think tank, who analyzed several hundred tenders issued from the PLA Procurement Network over six months this year.
Reuters could not independently confirm Cheung’s assertion, but the shift would coincide with a public pressure campaign by Beijing on domestic firms to use China-made technology.
The news agency’s review of procurement notices and patents filed to China’s patent office found demand for and use of Huawei chips by PLA affiliates, but it was not able to verify all the tenders seen by Jamestown, which on Monday released a report titled “DeepSeek use in PRC military and public security systems” — referring to the People’s Republic of China — that it provided early to Reuters.
Huawei declined to comment when asked about military deployment of its chips.
The Chinese Ministry of National Defense, DeepSeek and Norinco did not return requests for comment about their use of AI for military applications.
The universities and defense firms that filed the patents and research papers seen by Reuters also did not respond to similar questions.
Use of DeepSeek models was indicated in a dozen tenders from PLA entities filed this year and seen by Reuters, while just one referenced Alibaba’s Qwen, a major domestic rival.
Alibaba did not return a request for comment about military use of Qwen.
DeepSeek-related procurement notices have accelerated throughout this year, with new military applications appearing regularly on the PLA network, the Jamestown report said.
DeepSeek’s popularity with the PLA also reflects China’s pursuit of what Beijing calls “algorithmic sovereignty” — reducing dependence on Western technology while strengthening control over critical digital infrastructure.
The US Department of Defense declined to comment about the PLA’s use of AI.
A US Department of State spokesperson said in response to Reuters’ questions that “DeepSeek has willingly provided, and will likely continue to provide, support to China’s military and intelligence operations.”
Washington would “pursue a bold, inclusive strategy to American AI technology with trusted foreign countries around the world, while keeping the technology out of the hands of our adversaries,” the spokesperson added.
China is looking at AI-powered robot dogs that scout in packs and drone swarms that autonomously track targets, as well as visually immersive command centers and advanced war game simulations, the documents said.
In November last year, the PLA issued a sci-fi-esque tender for AI-powered robot dogs that would scout together for threats and clear explosive hazards.
Reuters could not identify whether the tender was fulfilled.
China has previously deployed armed robot dogs from AI robotics manufacturer Unitree in military drills, according to images released in state media.
Unitree did not respond to queries about its PLA work.
The review of patents, tenders and research papers published in the past two years shows how PLA and affiliated entities are looking to AI to improve military planning, including developing technology to quickly analyze images taken by satellites and drones.
Researchers at Landship Information Technology, a Chinese company that integrates AI systems into military vehicles, including Norinco’s, said in a February white paper released to promote its services that its technology built on Huawei chips can rapidly identify targets from satellite imagery, while coordinating with radars and aircraft to execute operations.
The time taken for military planners to shift from finding and identifying a target to executing an operation has also been shortened by AI, according to Xian Technological University.
Researchers at the institute said in a summary of their findings released in May that their DeepSeek-powered system was able to assess 10,000 battlefield scenarios — each with different variables, terrain and force deployments — in 48 seconds.
Such a task would have taken a conventional team of military planners 48 hours to complete, they said.
Reuters could not independently verify the researchers’ claims.
Chinese military entities are investing in increasingly autonomous battlefield technology, the documents suggest.
Two dozen of the tenders and patents seen by Reuters showed the military attempting to integrate AI into drones so they can recognize and track targets, as well as work together in formations with little human intervention.
Beihang University, known for its military aviation research, is using DeepSeek to improve drone swarm decisionmaking when targeting “low, slow, small” threats — military shorthand for drones and light aircraft — according to a patent filing this year.
Chinese defense leaders have publicly committed to maintaining human control over weapons systems amid growing concern that a conflict between Beijing and Washington could lead to the unchecked deployment of AI-powered munitions.
The US military, which is also investing in AI, is aiming to deploy thousands of autonomous drones by the end of this year in what officials say is an attempt to counter China’s numerical advantage in uncrewed aerial vehicles.
Chinese defense contractors such as Shanxi 100 Trust Information Technology have touted in marketing materials their reliance on domestically produced components such as Huawei’s Ascend chips, which allow AI models to operate.
The firm did not respond to questions about its relationship with Huawei and the PLA.
Despite the move to domestic processors, Nvidia hardware continues to be frequently cited in research by military-affiliated academics, according to a review of patent filings from the past two years.
Reuters identified 35 applications referencing use of Nvidia’s A100 chips by academics at the PLA’s National University of Defense Technology (NUDT) and at the “Seven Sons” — a group of Chinese universities under US sanctions and with a history of conducting defense-related research for Beijing.
Those entities in the same time period filed 15 patents related to AI applications that cited Huawei Ascend hardware, which was designed as a substitute to Nvidia chips.
As recently as June, the PLA Rocket Force University of Engineering separately filed a patent for a remote-sensing target detection system, which it said used A100 chips for model training.
Chinese Senior Colonel Zhu Qichao (朱啟超), who leads a NUDT research center, last year told Reuters that US restrictions have affected their AI research “to some degree,” although they were determined to narrow the technological gap.
Nvidia’s Rizzo played down PLA demand for Nvidia’s hardware, saying that China “has more than enough domestic chips for all of its military applications.”
A Chinese diplomat’s violent threat against Japanese Prime Minister Sanae Takaichi following her remarks on defending Taiwan marks a dangerous escalation in East Asian tensions, revealing Beijing’s growing intolerance for dissent and the fragility of regional diplomacy. Chinese Consul General in Osaka Xue Jian (薛劍) on Saturday posted a chilling message on X: “the dirty neck that sticks itself in must be cut off,” in reference to Takaichi’s remark to Japanese lawmakers that an attack on Taiwan could threaten Japan’s survival. The post, which was later deleted, was not an isolated outburst. Xue has also amplified other incendiary messages, including one suggesting
Chinese Consul General in Osaka Xue Jian (薛劍) on Saturday last week shared a news article on social media about Japanese Prime Minister Sanae Takaichi’s remarks on Taiwan, adding that “the dirty neck that sticks itself in must be cut off.” The previous day in the Japanese House of Representatives, Takaichi said that a Chinese attack on Taiwan could constitute “a situation threatening Japan’s survival,” a reference to a legal legal term introduced in 2015 that allows the prime minister to deploy the Japan Self-Defense Forces. The violent nature of Xue’s comments is notable in that it came from a diplomat,
Before 1945, the most widely spoken language in Taiwan was Tai-gi (also known as Taiwanese, Taiwanese Hokkien or Hoklo). However, due to almost a century of language repression policies, many Taiwanese believe that Tai-gi is at risk of disappearing. To understand this crisis, I interviewed academics and activists about Taiwan’s history of language repression, the major challenges of revitalizing Tai-gi and their policy recommendations. Although Taiwanese were pressured to speak Japanese when Taiwan became a Japanese colony in 1895, most managed to keep their heritage languages alive in their homes. However, starting in 1949, when the Chinese Nationalist Party (KMT) enacted martial law
“Si ambulat loquitur tetrissitatque sicut anas, anas est” is, in customary international law, the three-part test of anatine ambulation, articulation and tetrissitation. And it is essential to Taiwan’s existence. Apocryphally, it can be traced as far back as Suetonius (蘇埃托尼烏斯) in late first-century Rome. Alas, Suetonius was only talking about ducks (anas). But this self-evident principle was codified as a four-part test at the Montevideo Convention in 1934, to which the United States is a party. Article One: “The state as a person of international law should possess the following qualifications: a) a permanent population; b) a defined territory; c) government;
RSS