The White House accused Chinese entities on Thursday of conducting large-scale clandestine campaigns to copy US artificial intelligence models.
Michael Kratsios, the White House’s technology advisor, stated on X (formerly Twitter) that foreign entities, primarily from China, are engaging in “industrial-scale distillation campaigns” to steal US AI systems.
“Distillation” involves training an AI model to mimic the capabilities of a more powerful model by learning from its responses. While this practice is legal with permission, it is illegal when done secretly.
In late February, US-based AI company Anthropic accused three Chinese labs — DeepSeek, Moonshot AI, and MiniMax — of creating over 24,000 fraudulent accounts. These accounts allegedly initiated more than 16 million interactions with Anthropic’s AI model Claude to reverse engineer its functionality and train their own systems.
Similarly, OpenAI had previously accused DeepSeek of exploiting advanced circumvention techniques to secretly copy its AI models. This accusation was outlined in a memo the company sent to the US Congress on 12 February.
Kratsios revealed that unnamed foreign entities use tens of thousands of proxy accounts to avoid detection while using sophisticated evasion methods to extract US technological advances. He shared a memo with federal agencies detailing these concerns.
He also criticised the reliability of AI models built on such unethical foundations, suggesting their producers lack confidence in their integrity and dependability.
The Trump administration announced plans to share intelligence with US AI companies and explore potential measures against the responsible entities. However, no specific actions have been disclosed.
This announcement comes just weeks ahead of a scheduled summit between Donald Trump and Chinese President Xi Jinping in Beijing on 14 May.
