Beijing, 31 July 2026 – Chinese military-linked researchers have reportedly used US-developed artificial-intelligence models to help train defence-related systems, exposing the widening gap between national-security restrictions and the global accessibility of advanced software. The finding adds a new layer of risk for technology companies whose products can be repurposed far beyond their intended markets.

Unlike specialised hardware, AI models can be accessed through cloud services, software interfaces, published research and open-weight releases. That makes control enforcement more complex than blocking a shipment of advanced chips. Even when providers prohibit military or sanctioned use, researchers may route activity through intermediaries, local deployments or adapted versions that are difficult to trace.