White House accuses China of industrial-scale AI model distillation, commits to share intelligence with US companies

The White House Office of Science and Technology Policy released a policy memo on Wednesday accusing China of conducting 'industrial-scale' theft of American AI models through distillation. Michael Kratsios, director of OSTP, stated the US 'has evidence that foreign entities, primarily in China, are running industrial-scale distillation campaigns to steal American AI.' OpenAI accused DeepSeek in February of distilling its models; Anthropic on 23 February named DeepSeek, MiniMax, and Moonshot AI as having created approximately 24,000 fraudulent accounts that generated more than 16 million exchanges with Claude.

Distillation works by feeding thousands or millions of queries to a frontier AI model (advanced AI trained on large datasets), collecting the responses, and using those responses to train a cheaper rival model that approximates the original's capabilities. It does not require stealing model weights or breaking into servers. OpenAI's terms of service explicitly prohibit using outputs to create 'imitation frontier AI models,' but the legal status of distillation itself remains unsettled.

The memo directs federal departments to share intelligence with US AI developers about foreign distillation attempts and explore accountability measures, though no specific sanctions or enforcement actions were announced. Congress introduced the Deterring American AI Model Theft Act of 2026 on 15 April, which would direct the government to identify entities using 'improper query-and-copy techniques' and impose sanctions through the Commerce Department blacklist. The memo arrives three weeks before a planned Trump-Xi summit in Beijing on 14 May.