The growing artificial intelligence rivalry between the United States and China has brought a technical process known as “distillation” into the centre of a geopolitical dispute.
American AI companies, including Anthropic and OpenAI, have accused Chinese firms of using distillation to extract capabilities from their advanced AI models. In September, US agencies also accused six Chinese companies — DeepSeek, Moonshot AI, Alibaba, MiniMax, StepFun and Z.AI — of conducting large-scale distillation campaigns targeting American AI systems.
Distillation is not inherently an illegal or unusual AI technique. It involves using the outputs of a larger, more expensive model to help train or improve another model. The approach can allow developers to create capable systems with less computing power and training costs.
The controversy concerns how the technique is being used. US agencies allege that some Chinese companies systematically queried American AI models at large scale and extracted capabilities including reasoning, coding and other functions. Anthropic has separately reported what it described as major distillation campaigns involving Chinese AI companies.
China has rejected the US accusations. Chinese officials have described the claims as unfounded and defended the country's progress in AI as the result of domestic technological development.
The dispute has become part of the broader US-China competition over AI leadership, as both countries seek to develop increasingly capable models while governments debate technology controls, intellectual property and national security.
