|
Listen to article
|
The White House has declared that Chinese entities are running coordinated, industrial-scale campaigns to steal American artificial intelligence technology — and that it intends to fight back by working more closely with US companies to identify, expose and ultimately hold accountable the foreign actors exploiting their systems.
The warning came in an internal memo from Michael Kratsios, Director of Science and Technology Policy, who wrote that the administration had obtained new information confirming that “foreign entities, principally based in China” were systematically copying American AI advances through a process called distillation. The technique involves operating thousands of individual accounts on AI platforms — appearing to function as ordinary users — while coordinating attempts to extract proprietary information about how the underlying models work, then applying that extracted knowledge to build and train competing systems.
“The aim,” Kratsios wrote, was to “systematically undermine American research and development and access proprietary information.” He outlined four responses the White House would pursue: sharing more intelligence with US AI companies about the actors and tactics involved; better coordinating with those companies to counter the attacks; developing best practices for identifying, mitigating and remediating distillation campaigns; and exploring how the administration can hold foreign actors accountable.
Read also:
The memo named no specific entities and detailed no concrete enforcement actions. The White House declined to comment beyond its contents.
China’s embassy in Washington pushed back with vigour. A representative called the memo an example of “unjustified suppression of Chinese companies by the US” and rejected the framing of theft entirely. “China is not only the world’s factory but is also becoming the world’s innovation lab,” the representative said. “China’s development is the result of its own dedication and effort as well as international cooperation that delivers mutual benefits.”
Read also: Major NYC Storm Could Trigger $100B Insurance Loss — KCC
The companies that have actually experienced distillation activity have been less diplomatic in their descriptions of it. Anthropic earlier this year said it had identified distillation attacks from three Chinese AI laboratories — DeepSeek, Moonshot and MiniMax — finding all three had been working to copy Anthropic models through coordinated exploitation campaigns. OpenAI has made similar accusations against DeepSeek specifically. None of the three Chinese laboratories responded to requests for comment.
DeepSeek is the most prominent name on that list. The Hangzhou-based startup upended the global AI industry when it released its R1 reasoning model last year, claiming it had been developed for a few million dollars — a fraction of the hundreds of billions that American competitors like OpenAI, Google and Microsoft have been spending. The announcement triggered a sharp sell-off in AI-related stocks as investors recalculated whether the massive capital expenditures of US firms were as defensible as assumed. DeepSeek subsequently became one of the world’s most widely used AI models before suffering a major outage last month. The company is expected to release a new version imminently.
Read more:
The distillation accusation, if accurate, would reframe DeepSeek’s cost advantage in a way that carries significant implications for how the global AI race is understood. The American narrative has been that Chinese labs are building capable models on the cheap because they are exploiting research that US companies paid for — not because they have discovered more efficient methods independently. The Chinese counter-narrative is that the efficiency gains reflect genuine engineering innovation and that accusations of theft are a political instrument designed to suppress competition.
Read also: Trump Labor Chief Chavez-DeRemer Leaves White House
Kratsios acknowledged in the memo that defensive measures were improving. “As methods to detect and mitigate industrial-scale distillation grow more sophisticated, foreign entities who build their AI capabilities on such fragile foundations should have little confidence in the integrity and reliability of the models they produce,” he wrote — a warning directed as much at the quality of stolen technology as at the act of stealing it.
The memo arrives as Trump is expected to visit China in May, with AI alongside trade and the Iran war likely to feature prominently in any agenda. The administration has been escalating its technology competition posture toward Beijing across multiple fronts simultaneously — export controls on advanced chips, restrictions on Chinese technology in US infrastructure, and now a formal declaration that AI theft is a national security concern requiring a coordinated federal response.
Whether the four steps Kratsios outlined produce meaningful results, or whether they represent the kind of policy announcement that precedes action without quite constituting it, will depend on what the White House is actually prepared to do when it identifies the specific actors it has declined, so far, to name.
