US restrictions failed to stop China from using American AI to strengthen its military, as Beijing expands arms sales across Africa

Summary

Chinese military researchers affiliated with the People’s Liberation Army have drawn on AI models developed by OpenAI, Meta and Anthropic for defence-related studies, according to a Reuters investigation. Access occurred indirectly through publicly released model weights and distillation methods that transfer reasoning patterns from the original systems, rather than through direct commercial channels. No evidence indicates that the US companies knowingly supplied the PLA.

The work documented in more than sixty papers covers surveillance, cyber operations and tactical decision support. US export controls have focused on advanced chips such as Nvidia’s H20 series, yet the findings show that software artefacts and open-weight releases continue to reach Chinese defence laboratories. US officials have highlighted the potential of such models to improve intelligence analysis, battlefield planning and autonomous systems, underscoring the limits of hardware-centric restrictions.

At the same time, China has broadened its military engagement across Africa. It now ranks among the continent’s leading suppliers of aircraft, drones, armoured vehicles and air-defence equipment, with major recipients including Nigeria and Algeria. Training programmes, joint exercises and a permanent base in Djibouti further extend Beijing’s operational reach near key maritime routes.

Why it matters

This article is highly relevant for defense strategists and technologists as it highlights the limitations of hardware export controls and demonstrates how adversaries leverage open-source Western AI models for military applications. Understanding these dynamics is crucial for NATO and European defense professionals developing AI doctrines and security protocols.