China’s Military AI Faces a Critical Gap — Researchers Say Its “Attack Chain” Still Trails Foreign Models

China is rapidly pushing artificial intelligence into its military, but researchers inside the country say the People’s Liberation Army’s AI capabilities still face a major weakness: its systems have yet to match foreign models in performance, particularly when it comes to offensive combat applications.

A study by researchers affiliated with the Armed Police Force Shandong Corps and the Engineering University of the People’s Armed Police, published in China’s Defence Industry Conversion, says military AI applications remain concentrated largely on supporting functions such as intelligent question-answering and content generation. The researchers said these systems still “lag considerably” behind foreign large language models, while AI use in actual offensive operations remains at an exploratory stage. 

The findings are significant because Beijing has made AI a central part of its long-term military modernization strategy.

China wants AI to transform the battlefield

Chinese President Xi Jinping has repeatedly called for greater military use of AI and unmanned technologies.

In July, Xi urged the PLA to strengthen the military application of unmanned intelligent technologies, expand networked information systems and gradually build an “intelligent military system.” 

China has also recently revised its national defence mobilisation law to specifically promote the use of advanced technologies in defence mobilisation. Analysts say the provision could encompass AI, drones, cyberspace, quantum technology and other emerging capabilities. The revised law takes effect October 1, 2026. 

That makes the latest research particularly revealing: China is expanding the institutional framework for military AI even as researchers acknowledge that important technical gaps remain.

The “attack chain” problem

The challenge is not simply whether the PLA can build an AI model.

Modern military AI requires an entire chain connecting intelligence collection, battlefield awareness, decision-making, targeting, weapons and assessment.

Chinese researchers say offensive applications remain much less developed than support functions.

That distinction matters.

An AI system capable of generating text or answering questions is very different from one that can reliably process rapidly changing battlefield information and support complex military decision-making.

China itself has acknowledged the difficulty.

In August, Chinese state television reported on an AI-enabled “intelligent strike planning system” designed to help coordinate exercises involving hundreds of targets and dozens of aircraft formations. The system reportedly assists with target prioritisation, strike-wave coordination and task allocation. 

The two developments illustrate a complicated picture: China is clearly experimenting with sophisticated military AI, but its own researchers say the broader AI attack chain is not yet mature.

China is also borrowing from foreign AI

Another Reuters investigation adds an important layer to the story.

A review of more than 80 Chinese academic papers and patents found that researchers linked to the PLA and other military institutions have used outputs from U.S. AI models developed by companies including OpenAI and Anthropic to train smaller, specialised Chinese systems.

The technique, known as model distillation, transfers selected capabilities from a larger AI model into a smaller model that can operate with fewer computing resources. Reuters identified applications involving surveillance, cyber operations, drones and tactical decision-making. 

One documented example involved researchers from PLA Unit 96941 using GPT-3.5 to help process sensitive military software code and then training a domestic model using summaries generated from the U.S. system. 

However, experts caution that distillation does not simply reproduce the capabilities of the original frontier model. The resulting system generally inherits selected capabilities rather than the full breadth of its “teacher” model. 

But China is not standing still

The latest assessment should not be interpreted as meaning that Chinese military AI is ineffective or that Beijing is abandoning the technology.

Quite the opposite.

A February 2026 study by Georgetown University’s Center for Security and Emerging Technology examined thousands of PLA requests for proposals issued between 2023 and 2024.

The research found that the PLA is pursuing AI across command and control, communications, cyber operations, intelligence, surveillance, reconnaissance and targeting.

The objective is to use AI to process increasingly large amounts of information, accelerate decision-making and improve operational precision. 

Other Chinese research has also explored AI agents for battlefield command. A study published earlier this year described an AI system affiliated with the National University of Defense Technology that had reportedly been integrated into a command information platform supporting battalion-level operations. 

The U.S.-China AI military race is accelerating

The developments come as Washington is also moving aggressively to integrate AI into national security.

U.S. Strategic Command officials have described AI as increasingly important for sensing, analysis, planning, logistics and command and control, warning that countries capable of turning AI into faster decision-making could gain a major military advantage. 

The United States is simultaneously dealing with its own debate over how far AI should be allowed to go in warfare, including disputes over autonomous weapons and military surveillance. 

That means the competition is no longer simply about who has the best chatbot.

It is increasingly about who can turn AI into a reliable military system capable of processing information, supporting commanders and connecting sensors, personnel and weapons faster than an opponent.

What the latest Chinese research really tells us

The most important takeaway is not that China is “losing” the military AI race.

Rather, the research provides a rare glimpse into where the PLA believes its own weaknesses remain.

China has invested heavily in AI, drones, data systems and intelligent warfare. It is building the legal and technological infrastructure to expand those capabilities.

But according to Chinese researchers themselves, the transition from experimental and support-oriented AI to highly reliable offensive military applications remains unfinished.

And that gap could become one of the most important battlegrounds in the accelerating U.S.-China technology competition.

WWC ONE MEDIA J.M.D