The White House supports claims to have evidence that multiple entities in China are conducting large-scale campaigns to discreetly copy their American competitors and advancements in the AI sector.
The rivalry between China and the United States in the AI sector has escalated. The White House accused Chinese entities on Thursday of conducting “industrial-scale” campaigns to clandestinely copy American artificial intelligence models.
“The United States has evidence that foreign entities, primarily in China, are conducting distillation campaigns on an industrial scale to steal American AI,” said White House technology advisor Michael Kratsios.
“Distillation” is a technique that involves training an AI model on the responses of a more powerful model to replicate its capabilities. It is legal when authorized and illegal when conducted clandestinely.
Accusations deemed “completely unsubstantiated”
“These foreign entities,” not named by the White House, “use tens of thousands of proxy accounts” to evade detection “and circumvention techniques to systematically extract American technological advancements,” added Michael Kratsios, as he publicly released the memo he addressed to American federal agencies on Thursday.
“The foreign entities building on such fragile foundations must have little confidence in the integrity and reliability of the models they produce,” he quipped.
The Trump administration announced it would share its intelligence with American AI companies and “explore measures” against those responsible, without specifying which ones. This memo is made public less than three weeks before the planned summit between Donald Trump and his Chinese counterpart Xi Jinping in Beijing on May 14.
It also comes a few months after companies themselves criticized Chinese companies engaging in such practices. In late February, Anthropic accused three Chinese labs – Deepseek, Moonshot AI, and Minimax of creating “over 24,000 fraudulent accounts to generate more than 16 million exchanges with its Claude model,” to rebuild its functionality and train their own models.
A few days earlier, OpenAI accused Deepseek of clandestinely copying its AI models through sophisticated circumvention techniques in a memo addressed to the American Congress.
Unsurprisingly, the White House’s accusations did not sit well with Beijing. “The U.S. allegations are completely unsubstantiated. They constitute a slanderous defamation campaign against the successes of the Chinese AI industry,” criticized a spokesperson for the Chinese Foreign Affairs Ministry, Guo Jiakun, at a regular press briefing.
“We urge the American side to respect the facts, abandon its biases, cease its technological containment policy and repression towards China, and do more to facilitate technological exchanges and cooperation between the two countries,” he added, referring to the restrictions imposed by the United States on China’s access to certain technologies.
