The US accuses Chinese AI companies of copying American models at industrial scale

The United States has accused several Chinese AI companies of systematically using American AI systems to help train competing models.

US officials say companies including DeepSeek, Moonshot AI and Alibaba used variants or outputs of models developed by American companies such as OpenAI, Anthropic, Google and xAI.

The technique at the center of the dispute is model distillation.

Distillation is a widely used machine learning method in which a smaller model learns from the outputs of a larger or more capable model. The technique itself is not inherently malicious and is commonly used inside AI research.

The dispute concerns how the training data was obtained, whether platform rules were bypassed and whether large-scale extraction of model outputs effectively allowed competitors to reproduce capabilities developed by another company.

US officials described the alleged activity as malicious copying and argued that it could accelerate Chinese AI research while also strengthening military and cybersecurity capabilities.

The allegations arrive during a broader technological competition between the United States and China over advanced chips, AI models, infrastructure and access to computing resources.

Chinese AI companies have become increasingly competitive despite US restrictions on exports of advanced Nvidia hardware.

Distillation has therefore become more than a technical method. It is becoming part of a much larger debate over whether the capabilities of frontier AI models can be protected at all once users are allowed to interact with them.

The allegations are claims made by US officials and should not be treated as a judicial finding of wrongdoing.