AI integration by China's PLA marks shift in intelligence, cognitive warfare strategy

By Li Hsian | 2025-07-18

In recent years, China's People's Liberation Army (PLA) has actively invested in generative artificial intelligence (AI) for applications ranging from intelligence analysis to influence operations, according to researchers.

The PLA and China's defense industry are integrating both domestic and foreign large language models (LLMs) to develop generative AI tools for military and intelligence use, according to a report published June 17 by Insikt Group, cybersecurity firm Recorded Future's threat research division.

The PLA has already designed methods and systems that utilize generative AI for core intelligence tasks such as producing open-source intelligence (OSINT) products, processing satellite imagery, extracting events and handling intelligence data, according to the report.

Chinese defense contractors have claimed to have provided PLA units with tools built on domestic models like DeepSeek, it added. The PLA likely adopted DeepSeek rapidly after the release of its V3 and R1 models last December and in January, respectively.

Chinese patent filings reviewed by Insikt Group and Reuters show that state-affiliated research institutes, including the Academy of Military Sciences, have proposed training military LLMs using multisource intelligence inputs, including OSINT and human, signal, geospatial and technical intelligence.

These models are designed to support all phases of the intelligence cycle and improve battlefield decision-making.

Notably, some proposals suggested using OpenAI's Sora video generation model for cognitive warfare, including the creation of fake but realistic imagery to influence adversary decisions.

"Chinese covert influence networks have used generative AI to carry out online influence operations," the Insikt Group report said, indicating that such capabilities have moved from experimental phases into real-world operations.

China’s AI infrastructure expansion is raising concerns.

China has established at least one data center in every province, a joint May report by Strider Technologies and the Special Competitive Studies Project found.

By mid-2024, China had built or announced more than 250 AI data centers to reach its 2025 target of 105 EFLOPS (a measure of computing power) in AI computing. The total projected computing capacity could exceed 750 EFLOPS across all workloads.

The report identifies two overseas facilities, including one operational in Jakarta since May 2024 and another announced in early 2025 in Pasig, the Philippines, as part of China's effort to expand global computing capacity beyond US export controls.

High stakes

If China integrates generative AI into intelligence analysis and military planning, the PLA could gain superior situational awareness in the Indo-Pacific, particularly in the Taiwan Strait, South China Sea and East China Sea, Wang Xiu-wen, assistant research fellow at the Institute for PRC Military Affairs and Operational Concepts at Taiwan's Institute for National Defense and Security Research (INDSR), told Focus, an affiliate publication of Global Watch.

China's AI-assisted decision-making appears increasingly sophisticated, she said.

Systems such as "Zhàn Lú" (War Skull) can autonomously analyze battlefield inputs, issue commands and select optimal weapon combinations, allowing faster and more automated warfare led by unmanned platforms.

However, current generative AI remains prone to serious factual errors, said Wang.

In high-stakes regions, such mistakes could escalate into unintended conflicts.

As the PLA accelerates the militarization of generative AI, the international security landscape faces mounting challenges.

A 2023 report by the Center for a New American Security observed: "The unfortunate but inescapable reality is that East Asia is seeing a military-technological arms race, of which military AI is a part."

AI-enabled cognitive warfare and disinformation campaigns are further complicating the information environment in sensitive areas like the Indo-Pacific, lowering the threshold for confrontation.

To address these threats, the Insikt Group report recommends that the United States and its allies closely monitor developments in PLA AI adoption, assess risks related to technology transfer and to counterintelligence and formulate strategic responses.

Wang likewise urged democratic nations in the Indo-Pacific to accelerate intelligence sharing and integration and to explore technologies such as blockchain for decentralized information storage. Doing so could enhance agility and resilience in countering China's AI-driven military intelligence systems.