US AI Giants Accuse Chinese Labs of Industrial-Scale Model Distillation

Major U.S. artificial intelligence companies are accusing their Chinese rivals of using a technique called "distillation" to replicate their chatbots at industrial scale.

Anthropic disclosed that three Chinese labs-DeepSeek, Moonshot AI, and MiniMax-created roughly 24,000 fake accounts. These accounts conducted over 16 million conversations with Anthropic's Claude model, harvesting responses to train competing systems.

Distillation is a legitimate machine learning method. A smaller model learns to mimic a larger one. The issue arises when the "teacher" model belongs to someone else.

OpenAI has made parallel accusations against DeepSeek. In June 2026, new allegations targeted Alibaba in what is described as the largest known distillation operation, involving approximately 28.8 million interactions with AI models.

In response, OpenAI, Anthropic, and Google began sharing threat intelligence in April 2026. A White House memo confirmed the administration views these campaigns as "industrial-scale" operations by Chinese companies.