Back to News
SecurityAI Understanding briefing

US report says Chinese military employs commercial AI models for cyber training

A US congressional report reveals that China’s People’s Liberation Army is using commercial AI models such as DeepSeek and Alibaba’s Qwen for cyber‑range exercises and other military applications.

4 min readRead the linked source
Source-provided image accompanying US report says Chinese military employs commercial AI models for cyber training
Source referenceSource recorded
Publisher
lokmattimes.com
Source type
Linked source — primary-source status has not been established.
ContextUnderstand this in 60 seconds

Key terms

Natural Language Processing (NLP)
The branch of AI focused on understanding and generating human language.
AI Governance
Policies, standards, and oversight mechanisms that guide how AI is developed and used in society.
Prompt
The input instructions and context provided to a generative model.
Test yourselfAI Ethics Quiz

What happened

A bipartisan House report disclosed that the People’s Liberation Army (PLA) has procured servers to run commercial large‑language models—including DeepSeek, Alibaba’s Qwen, Moonshot AI’s Kimi and MiniMax—within isolated networks for cyber‑training, software development and AI‑agent tasks. The report cites procurement records and notes multiple visits by members of a PLA cyber‑intelligence unit to Zhipu AI’s Beijing headquarters, suggesting a direct liaison. Lawmakers are calling for expanded U.S. export controls and labeling requirements to block Chinese‑origin AI from U.S. federal contracts and critical infrastructure.

The House Select Committee on China and the House Homeland Security Committee compiled a report that cites procurement records showing the PLA’s request for high‑performance servers capable of running large‑language models. The report lists DeepSeek, Alibaba’s Qwen, Moonshot AI’s Kimi and MiniMax as models under evaluation.

According to the report, representatives of a PLA cyber‑intelligence unit visited Zhipu AI’s headquarters in Beijing multiple times, the latest visit occurring in August. While the exact content of those meetings was not disclosed, the frequency suggests an ongoing channel for technology transfer or collaboration.

The PLA’s stated objectives for using these models include "intelligent attack" and "intelligent penetration" scenarios in cyber‑range exercises, as well as non‑combat uses such as generating synthetic medical records and supporting clinical decision‑making. The models are reportedly being run on isolated internal networks to mitigate leakage of proprietary data.

U.S. lawmakers, led by Republican chairs John Moolenaar and Andrew Garbarino, are urging the administration to broaden restrictions that would bar Chinese‑origin AI models from U.S. federal contracts and critical infrastructure projects. They also propose a labeling standard so buyers can identify the underlying AI model in any purchased product.

Source details: lokmattimes.com

Why it matters

The integration of commercially available large‑language models into PLA cyber‑training marks a significant escalation in the militarization of AI, blurring the line between civilian AI products and state‑sponsored offensive capabilities. If Chinese forces can simulate sophisticated attacks and generate realistic medical data using these models, they may lower the technical barrier for advanced cyber operations and accelerate the development of autonomous decision‑making tools. The report’s findings also fuel policy debates in Washington about tightening export controls on AI hardware and software, and mandating transparency about the underlying models in procured systems. Such measures could shape the future of global and affect multinational AI supply chains.

The report highlights a concrete example of how commercial AI is being repurposed for military use, raising concerns about the speed at which adversaries can adopt cutting‑edge AI capabilities without developing them in‑house.

By leveraging off‑the‑shelf models, the PLA may reduce research and development costs while gaining access to state‑of‑the‑art natural language processing, code generation and data synthesis tools, potentially enhancing the sophistication of cyber‑espionage and cyber‑attack operations.

The findings could a reassessment of U.S. export control regimes, which currently focus more on hardware than on software and AI models. Introducing model‑level labeling could increase transparency but also impose compliance burdens on vendors.

If Chinese military units can effectively integrate these models into isolated networks, it may set a precedent for other nations to follow, accelerating a global arms race in AI‑enabled cyber capabilities.

Interactive Mechanism

Interactive Mechanism: How It Actually Works

Explore the underlying technology behind this development interactively.

Agent Lifecycle Stage:
1
User Intent & Planning: "Audit customer refund request #4092 and settle payment."
2
Tool Calling: Emits structured JSON call crm_get_transaction(id='4092').
3
Guardrail & Verification:Paused: High-value action requires human operator sign-off.
4
Final Settlement: Refund recorded, email receipt dispatched, and audit log stored.
Core takeaway: An AI agent is not just a language model—it is a closed loop of planning, tool invocation, and environment feedback. Production systems require self-healing retries and strict human approval guardrails.
Interactive Concept Check+10 Points
AI Ethics Quiz

Why can ethical evaluation not be reduced to one model score?

What to watch next

Future congressional hearings on AI export controls, potential sanctions on Chinese AI firms, and any official response from the Chinese Ministry of Defense or the companies named. Watch for U.S. Department of Commerce actions on AI‑related technology transfers and for Chinese policy shifts regarding the export of AI models abroad.

Upcoming hearings of the House Select Committee on China where the report’s authors may be questioned and further evidence could be released.

Potential legislative proposals to expand the Export Administration Regulations (EAR) to cover AI models and related software, as well as any executive actions from the Department of Commerce.

Official statements from the Chinese Ministry of Defense or the AI firms named (DeepSeek, Alibaba, Moonshot AI, MiniMax, Zhipu) regarding the allegations and any planned policy responses.

Industry reactions, particularly from U.S. cloud providers and AI hardware manufacturers, about the feasibility of model‑level labeling and its impact on supply chains.

Related guides & quizzes

Found this useful?