Skip to content
415601567c03a30af4ca499200fed35f8bbb5c42

AI Lessons From Russia’s War in Ukraine

Military & Security Publication China Brief China Volume 26 Issue 21

10.30.2026 Sunny Cheung

Download

AI Lessons From Russia’s War in Ukraine

Executive Summary:

  • Chinese military writers studying the use of artificial intelligence (AI) in Russia’s war in Ukraine have concluded that its decisive value lies in shortening the kill chain rather than in fully autonomous weapons.
  • In their reading, AI plays four roles: turning sensor data into targets, accelerating planning and command, keeping drones on target when jamming cuts the link, and mass-produced propaganda and deep-fakes.
  • Each of these roles rests on the combat data, computing power, chips, and AI models that the PRC wishes to control domestically. Starlink’s wartime restrictions and Russia’s reliance on U.S. chips serve as warnings.
  • The People’s Liberation Army (PLA) is turning these lessons into AI planning systems, unmanned units, and procurement of domestic large language models (LLMs). Some concede that PLA’s AI is still used mainly for auxiliary tasks, with these authors acknowledging that doctrine lags technological advancements.

Editor’s note: This article is the latest in an occasional series by Jamestown China Studies Fellow Sunny Cheung on Chinese lessons from Russia’s war in Ukraine.

On August 2, China Central Television (CCTV) aired an episode of Zhisheng (制胜; “Victory”), an Army Day documentary series, showing a People’s Liberation Army (PLA) Air Force “intelligent strike-planning system” that enabled more than 100 tactical units to conduct synchronized strikes (Global Times, August 3). In early October, troops from the PLA Central Theater Command began returning from Russia’s multinational Center 2026 exercise, where drone swarms fed precise target information to ground combat groups (Global Times, October 3). And on October 12, authors from the National Defense University’s Joint Operations College wrote in the PLA Daily that artificial intelligence (AI) has “leapt from a combat auxiliary tool to the core enabling engine of the command-support system” (从作战辅助工具跃升为指挥保障体系的核心赋能引擎), citing Ukraine’s Delta and Russia’s Svod battle-management systems as proof (PLA Daily, October 12).

These developments show that Chinese military experts have consistently learned from Russia’s war in Ukraine, which they see as the first major test of “intelligentized” (智能化) warfare—the PLA’s term for war in which AI pervades command and weapons systems. Chinese research papers published in military journals over the last two years make clear that AI’s principal contribution to military victory is in compressing the time lag in a kill chain from the initial order to executing a strike. It does this by assisting with four primary tasks: finding targets, making decisions, striking under jamming, and waging cognitive warfare.

Turning Data into Strikes

Chinese researchers are impressed by what Ukrainian writers call the “Didi ride-hailing” (滴滴打车) model, named after the PRC’s Uber equivalent, in which small units accept piecemeal strike orders via tactical apps and report back when they have been carried out. In September 2024, Ukraine’s Defense Ministry reported that its “Avengers AI” platform automatically detected roughly 12,000 Russian equipment targets per week via analysis of video from drones and fixed cameras, something that impressed researchers at the National University of Defense Technology (Ministry of Defense of Ukraine, September 23, 2024; Zhang and Liang, 2025). [1] A paper led by Fang Chao (房超), a professor at Huazhong University of Science and Technology and former chief strategy research scientist at Qiyuan Laboratory (启元实验室), contends that the GIS Arta artillery app, a Ukrainian targeting app that matches artillery units with targets reported by drones and spotters, has slashed the time between receiving a fire order and striking a target from 20 minutes to just 30 seconds (Fang Chao et al., 2025). [2] Researchers point to the whole process—from sifting through enormous volumes of drone and satellite feeds, to decoding intelligence gathered from intercepted communications and civilian tips, to the speed of target acquisition—as a model to emulate.

Experts from the National University of Defense Technology believe that the PLA should treat private sector AI and big data capabilities as part of its arsenal, cataloguing them as “potential resources” (潜力资源) for conversion into military uses (Zhang and Liang, 2025). [3] Li Chen (李晨) of Renmin University is more measured in his assessment, noting that although information-network operations and electronic warfare have figured heavily in the Ukraine conflict, they have not “paralyzed either side’s operational system” (未导致任何一方作战体系瘫痪) (Li Chen, 2025). [4] Ultimately, he argues, the advantage from faster target identification has not been decisive.

AI Making Decisions in the Command Post

Several experts believe that AI will be instrumental in improving decision-making processes in future wars. An article in the journal Military Digest (军事文摘), for example, argues for the creation of “joint operations intelligent computing centers” (联合作战智算中心) and specialized large models trained on joint operations scenarios to deliver “dimension reduction strikes” (降维打击). This refers to attacks from a level of capability so far above the opponent’s that it cannot respond in kind, which could be used against “the U.S.-allied intelligentized combat system” (美盟智能化作战体系). In a future high-end conflict, the author writes, “what is fought over is information, what is fought with is algorithms, and what is contested is intelligence” (斗的是信息,打的是算法,争的是智能) (Li Zhongqiang, 2025). [5]

The PLA has made progress in using gathered intelligence in the decision-making process. The Global Times reported that the Air Force strike-planning system seen on CCTV can coordinate targets, formations, and successive waves of attack far more quickly than conventional planning (Global Times, August 3). An aerospace editor at state-affiliated China Aerospace Daily reported the next day that intelligence gathered from PLA Air Force drones can be fed into an open-architecture system that relies on AI to evaluate it and “generate optimal operational plans in seconds or tens of seconds” (Global Times, August 4).

According to procurement records, PLA entities have issued tenders that explicitly call for tools based on DeepSeek and other domestically developed AI models, while university researchers use DeepSeek to examine battlefield scenarios and to improve drone-swarm targeting decisions (Reuters, October 27, 2025; China Brief, October 27, 2025). However, members of the People’s Armed Police (PAP), the paramilitary police force under the command of the Central Military Commission, have acknowledged that PLA AI applications remain limited to auxiliary support functions such as answering queries and generating content, and lag foreign models in performance considerably. The application of AI in offensive combat remains at an exploratory stage (Li, Wu, and Wang, 2026). [6] Experts at the PRC’s National Defense University see militaries moving toward a hybrid approach in which AI assists with various functions but humans make the final call (AI辅助推演、人类最终定夺)—something also referred to as “humans in the loop, machines speeding up” (人在回路、机在提速) (PLA Daily, October 12). Ukraine’s new decision-support system Marichka is an early version of this kind of system. The same PLA Daily article voices fears, however, that if AI is integrated too deeply into decision mechanisms, “human-machine authority will be eroded in the pursuit of efficiency in war” (人机权责在战争中被效率侵蚀) (PLA Daily, October 12).

Mixed Views on Autonomy’s Decisiveness

Autonomy addresses the problem of communications jamming. As Yang Zhen (杨震) of the China Electronics Standardization Institute wrote for the defense industry publication China Defense Conversion (中国军转民), “the core lesson of the Russia–Ukraine battlefield is that the electronic warfare environment determines the technology route” (俄乌战场的核心教训是电子战环境决定了技术路线), but fully autonomous use of lethal force is still far from being mainstream (Yang Zhen, accessed October 30, 2026). [7] Autonomy is nevertheless widely understood to be potent, and Chinese defense-industry researchers have been quick to take note of the use of autonomy in Ukraine. An engineer at the 54th Research Institute of the China Electronics Technology Group Corporation (CETC) mentioned that Ukrainian forces now routinely use “last-mile” autonomy, in which a drone can identify a target, hone in on it, and make impact after communications with the operator are jammed. “Last-mile” autonomy boosts hit rates in a strike mission from 10–20 percent to 70–80 percent, and troops can learn to use the technology in under 24 hours (Hao Yunfeng, April 2025). [8]

Some PLA writers see autonomy as fully decisive. One author in the Military Digest urged the PLA to build forces around algorithms, stating that “algorithms are king, the system wins” (算法为王、体系制胜), and using Ukraine’s June 2025 Operation Spiderweb as evidence that “human-out-of-the-loop” (人在回路外) operations are increasingly common. Despite the piece’s enthusiasm, however, it makes unsupported claims about AI swarm coordination, and PLA doctrine lags behind its proposals (Liu Zhenggang, 2026). [9] As authors from PLA Unit 63893 wrote in 2025, drone attack-and-defense theory in general “has not yet formed valuable combat guidance manuals or regulations” (尚未形成有价值作战指导手册、作战条例等) and called deeper research “a real and urgent task for us” (对我们来说是一项现实而紧迫的工作) (Liu Zhengtang et al., 2025). [10]

The PLA is instead focusing on building unmanned ground units and counter-drone capabilities. At the September 2025 military parade in Beijing, unmanned ground formations were described as capable of “remote control, autonomous action, and flexible grouping” (远程操控、自主行动、灵活编组) (Xinhua, September 5, 2025). In an exercise the following month, a PLA brigade trained counter-drone tactics, using equipment that could detect drones’ radio frequencies and match them to a catalog of more than 800 commercial drone models (CSIS, October 7).

Cognitive Warfare Offers Cheap Offensive Measures

The fourth PLA use case for AI is deception. Military researchers have worked on deepfake technologies with the aim of deceiving, persuading, and shaping the cognitive battlefield. Deepfake technology is cheap—researchers at a Chinese military academy note that they can create a voice clone by using just a ten-second audio sample and that the cost of creating deepfakes has collapsed to the point that “the side defending against deepfakes always pays more than the side using them” (防御深度伪造技术的一方付出的代价总是要多于应用它的一方) (Gu Junwei et al., 2025). [11]

Rather than curbing deepfakes through detection, the paper recommends “offense as defense” (以攻代守), or the pre-emptive creation of deepfakes by the defensive side. By creating a flood of its own deepfakes, an army can seize the “cognitive high ground” (抢占认知域的高地). Doing so, the authors argue, will save money and prove effective. Another paper cites an example of Russian attempts to create low-quality websites that, when scraped for data to train large language models, pollute the dataset (Wang Ling et al., 2026). [12] In this way, the authors argue, AI can be used to deceive the enemy, just as it is already used to deceive consumers.

There is some debate about how effective these techniques can be. Writing in the PLA Daily earlier this year, an academic from the Academy of Military Sciences argued that AI enables more granular targeting of propaganda, turning propaganda wars from “flood irrigation” (漫灌) into more precise “drip irrigation” (滴灌) (PLA Daily, April 13). By contrast, civilian scholars note that AI propaganda may not pay off. In Ukraine, they say, the sheer volume of AI-generated material led to massive fatigue and mistrust, and ultimately limited its influence (Shi and Huang, 2025). [13]

Fears of Hardware Dependencies

The AI technologies empowering modern military capabilities rely on an array of software and hardware supply chains that are vulnerable. Chinese experts treat control of these resources as an essential military requirement. First, AI needs combat data for training. According to a Military Digest assessment of Russian military AI, Moscow started collecting combat data from unmanned combat systems in mid-2025 to train its AI models, which cannot be substituted by laboratories. Li Zhongqiang (李忠强), a PLA researcher, has called for “an all-army mechanism for reporting military training data samples” (全军军事训练数据样本上报机制) (Li, 2025). [14] Second, AI needs hardware. The same Military Digest assessment notes that Nvidia’s Jetson Orin Nano, an edge AI computer, is used in Russia’s long-range Shahed drone, the MS001, while more than half of the key AI components in Russia’s unmanned systems are supplied by U.S. firms (Yuan Li et al., 2026). [15]

Third, AI needs networking. As Peking University researcher Pang Xun (庞珣) points out, SpaceX can unilaterally control the coverage area and functional scope of a sovereign state’s critical wartime infrastructure by deciding where its Starlink system is available (Pang and Qu, 2026). [16] Gao Jixiang (高际香), a researcher at the Chinese Academy of Social Sciences, draws the conclusion from Russia’s experience of Western sanctions that only self-controlled AI can ensure the continuation and iteration of a nation’s AI system in the face of a blockade (Gao and Liu, 2026). [17] Moscow has thus turned to its Chinese partners for assistance, with Sberbank cooperating on the “joint improvement and adaptation of open-source large language models” (开源大语言模型的共同改进和适配) (S&T Daily, September 25, 2025). This assistance would be in conflict with Russia’s draft AI law, however, which excludes foreign software or components from its definition of “sovereign” models (Gao and Liu, 2026). [18] Clearly, sovereignty is easier to claim than to achieve.

Conclusion

Chinese military writers interpret Russia’s war in Ukraine as one in which AI matters most as an accelerator—of targeting, planning, and strikes under jamming—and as a cheap amplifier of propaganda. They also observe that all of these innovations rest on data, chips, and networks that must be brought under national control. They are divided, however, on some of the other takeaways from the conflict. Some argue that the PLA should develop AI planning systems, unmanned formations, and domestically-produced large language models, while others argue that the PLA should focus instead on developing “upstream capabilities” (上游能力建设) such as intelligence, navigation, communications, and command-and-control, as well as operator training, arguing that these are better investments than drones (Huang and Yan, 2026). [19]

The establishment of a system-wide method for collecting training data, formation of special units to employ unmanned systems, and exercises designed to test unmanned systems in heavily-jammed environments are all worth watching out for in the coming months. After all, PLA officers are aware of the shortcomings of PLA doctrine, performance of AI models, and lack of combat experience, and Beijing wants to understand the war in Ukraine before having to fight its own.

The author would like to thank the Smith Richardson Foundation for supporting this work.

Notes

[1] Zhang Chengxian [张成贤] and Liang Tian [梁天]. “Characteristics of Ukrainian Battlefield Target-Intelligence Support in the Ukraine Crisis” [乌克兰危机中乌军战场目标情报保障特点研究]. In Proceedings of the 13th China Command and Control Conference [第十三届中国指挥控制大会论文集], 153–57. 2025. Both authors are at NUDT’s College of International Studies.

[2] Fang Chao [房超]. “Typical Features and Winning Rules of Intelligentized Warfare as Seen in the Russia–Ukraine Conflict” [从俄乌冲突看智能化战争的典型特征与制胜规律]. China Defense Conversion [中国军转民], no. 10 (2025): 14–17. Unattributed think-tank analysis republished in a defense industry magazine.

[3] Zhang Chengxian [张成贤] and Liang Tian [梁天]. “Characteristics of Ukrainian Battlefield Target-Intelligence Support in the Ukraine Crisis” [乌克兰危机中乌军战场目标情报保障特点研究]. In Proceedings of the 13th China Command and Control Conference [第十三届中国指挥控制大会论文集], 153–57. 2025. Both authors are at NUDT’s College of International Studies.

[4] Li Chen [李晨]. “Technological Innovation, AI, and Great-Power Military Competition” [技术创新、人工智能与大国军事竞争]. Journal of National Security Studies [国家安全研究], no. 1 (2025): 46–64. Li is vice dean of Renmin University’s School of International Studies.

[5] Li Zhongqiang [李忠强]. “The Use of AI Algorithms in Joint Operations and Reflections” [人工智能算法在联合作战中的运用与思考]. Military Digest [军事文摘], no. 3 (2025): 36–39.

[6] Li Xinfu [李新福], Wu Xingfang [吴星芳], and Wang Xiao [王啸]. “Research on AI-Driven Offensive Military Applications” [基于人工智能驱动的攻击性军事应用研究]. China Defense Conversion [中国军转民], no. 14 (2026): 45–46. The authors are with the PAP Shandong Corps and the PAP Engineering University. The paper concerns U.S. cyber operations rather than Ukraine; its account of those operations contains claims not supported by open sources.

[7] Yang Zhen [杨震]. “AI Research in U.S. Defense: From U.S. Policy to Combat Experience on the Russia–Ukraine Battlefield” [美国国防领域的人工智能研究——从美方政策到俄乌战场的实战经验]. China Defense Conversion [中国军转民], “Focus” section, n.d., 18–20. (issue number not printed on the pages). The China Electronics Standardization Institute is subordinate to the Ministry of Industry and Information Technology.

[8] Bondar, Kateryna. “Ukraine’s Future Vision and Current Capabilities for Waging AI-Enabled Autonomous Warfare (Translation)” [乌克兰当前人工智能自主作战能力与未来愿景 (译文)]. Abridged translation by Hao Yunfeng [郝云丰]. Information Security and Communications Privacy [信息安全与通信保密], no. 4 (2025): 17–25. Originally published by the Center for Strategic and International Studies (CSIS). The translator is an engineer at CETC’s 54th Research Institute; the CSIS original was published in March 2025, one month before the translation.

[9] Liu Zhenggang [刘政刚]. “Lessons and Strategic Thinking from Ukraine’s ‘Operation Spiderweb’ for Intelligentization” [乌克兰‘蛛网行动’对智能化建设的启示与战略思考]. Military Digest [军事文摘], no. 4 (2026): 18–23.

[10] Liu Zhengtang [刘正堂], Zou Yongjie [邹永杰], Sun Jian [孙健], and Guo Enze [郭恩泽]. “Analysis of and Lessons from UAV Attack and Defense in the Russia–Ukraine War” [俄乌战争中无人机攻防运用分析与启示]. Fire Control & Command Control [火力与指挥控制] 50, no. 10 (2025): 207–12. The authors are from PLA Unit 63893, Luoyang. The paper also misdates the drone attack on the Kremlin to May 2024 (it occurred in May 2023).

[11] Gu Junwei [顾俊伟], Kong Zhiqi [孔祉麒], and He Kang [何康]. “Deepfake Technology in Cognitive Confrontation: Applications and Reflections” [深度伪造技术在认知对抗中的应用与思考]. In Proceedings of the 13th China Command and Control Conference [第十三届中国指挥控制大会论文集], 603–8. 2025. The authors are at NUDT’s College of International Studies.

[12] Wang Ling [王凌], Zhao Zongxu [赵宗旭], and Bu Jiang [卜江]. “Weaponized Narratives: Social Bot Strategies in the Ukraine Crisis and Their Lessons” [武器化的叙事:社交机器人在乌克兰危机中的应用策略及其启示]. Military Digest [军事文摘], no. 5 (2026): 17–22.

[13] Shi Peipei [石培培] and Huang Chengchun [黄成春]. “A Comparison of AI-Driven Cognitive Warfare in the Russia–Ukraine and Israel–Palestine Conflicts” [人工智能驱动下俄乌冲突与巴以冲突的认知战比较]. Journal of Nanjing University of Posts and Telecommunications (Social Science Edition) 27, no. 6 (2025): 9–19. Shi is at the CASS Institute of American Studies.

[14] Li Zhongqiang [李忠强]. “The Use of AI Algorithms in Joint Operations and Reflections” [人工智能算法在联合作战中的运用与思考]. Military Digest [军事文摘], no. 3 (2025): 36–39.

[15] Yuan Li [袁理], Zhou Qi [周琦], and Yang Aihua [杨爱华]. “An Analysis of Russia’s Military AI Development” [俄罗斯军用人工智能发展透析]. Military Digest [军事文摘], no. 7 (2026): 20–24. The paper relies heavily on two CSIS reports from April 2026; several of its personnel-loss figures are unsourced.

[16] Pang Xun (庞珣) and Qu Mingzhu (曲明珠), “人工智能介入俄乌冲突对大国关系的影响” [The Impact of AI’s Involvement in the Russia–Ukraine Conflict on Great-Power Relations], Russian Studies (俄罗斯研究), No. 1 (2026), pp. 65–90. Pang is a professor at Peking University’s School of International Studies.

[17] Gao Jixiang [高际香] and Liu Dejian [刘德健]. “Strategic Choices and Dilemmas of Russian AI Development under Sanctions” [制裁下俄罗斯人工智能发展的战略选择与困境]. Russian, East European & Central Asian Studies [俄罗斯东欧中亚研究], no. 5 (2026). Gao is a researcher at the CASS Institute of Russian, East European and Central Asian Studies.

[18] Gao Jixiang [高际香] and Liu Dejian [刘德健]. “Strategic Choices and Dilemmas of Russian AI Development under Sanctions” [制裁下俄罗斯人工智能发展的战略选择与困境]. Russian, East European & Central Asian Studies [俄罗斯东欧中亚研究], no. 5 (2026).

[19] Huang Mingtuo [黄明拓] and Yan Guodong [阎国栋]. “The Changing Form of War with the Intervention of Unmanned Weapons” [无人武器介入下的战争形态变迁]. Journal of Yanshan University (Philosophy and Social Science Edition), no. 5 (2026): 68–80. Both authors are at Nankai University’s School of Foreign Studies.

Jamestown
Privacy Overview

This website uses cookies so that we can provide you with the best user experience possible. Cookie information is stored in your browser and performs functions such as recognising you when you return to our website and helping our team to understand which sections of the website you find most interesting and useful.