White House says China is stealing AI know-how from US labs

In short: The White House says China-linked groups are running large-scale efforts to copy advanced US AI systems, and it plans a crackdown.

What happened

The White House has accused foreign groups, mainly based in China, of stealing key ideas and methods from American artificial intelligence labs. Michael Kratsios, who leads the White House Office of Science and Technology Policy, made the claim in a memo reported by the Financial Times.

The memo focuses on a tactic called “distillation.” Distillation is a way to train a smaller AI model by studying the answers from a larger AI model (like a student copying the work of a top student to learn faster). Kratsios said the US government believes this is happening at “industrial scale,” meaning in a very large and organized way.

Kratsios said these efforts may use “tens of thousands” of proxy accounts, which are extra accounts used to hide who is really behind the activity. He also mentioned “jailbreaking,” which means trying to get an AI system to ignore its built-in rules so it reveals information it should not share.

The administration said it will share information with US AI companies about suspected attempts at unauthorised distillation and help companies coordinate their defenses. It also said it would explore measures to hold foreign actors accountable.

US AI companies including Anthropic and OpenAI have raised concerns about distillation before. Anthropic has accused DeepSeek, Moonshot, and MiniMax of distillation attacks, and OpenAI has said it had evidence that DeepSeek used outputs from GPT models against its rules.

Why it matters

If companies can cheaply copy powerful AI systems, it could reduce the advantage of labs that spend heavily on research. US firms also say copied models may remove safety checks, which could increase risks like misuse for hacking or dangerous biological research.

Source: Financial Times