White House Memo Details Alleged Chinese AI Distillation Campaigns
The White House has accused foreign entities, primarily based in China, of conducting coordinated campaigns to copy frontier AI models from US companies using thousands of surrogate accounts. The accusations involve distillation techniques that extract capabilities from American AI systems.
thehindu.comForeign entities primarily based in China have conducted industrial-scale campaigns to copy frontier AI models from US companies, according to a memo issued by the White House director of the office of science and technology policy. The director issued the memo on Thursday, stating that these campaigns entail using tens of thousands of surrogate accounts to go undetected.
The efforts involve complex tools to expose proprietary information.
These coordinated campaigns systematically extract capabilities from American AI models, exploiting American expertise and innovation, the director wrote. The White House stated that China is engaged in deliberate, industrial-scale campaigns to distill US frontier AI systems.
The government has learned that Chinese entities are leveraging thousands of proxy accounts. The accusations center on a process called distillation, a technique used to transfer knowledge from a large AI model to a smaller one that can run more cheaply. Distillation is widely used for training AI models.
Companies like Anthropic and OpenAI have previously claimed the distillation process has been used to unfairly mimic their models’ capabilities. DeepSeek, a Chinese AI startup, rattled Wall Street last year. DeepSeek has been at the center of accusations of using distillation on US models.
OpenAI warned lawmakers in February that DeepSeek has been attempting to replicate the performance of OpenAI’s models and those of other US frontier labs through distillation. Anthropic identified industrial-scale campaigns by DeepSeek along with two other AI labs to illicitly extract capabilities to improve their own models. Anthropic stated this identification occurred in February.
DeepSeek did not immediately respond to CNN’s request for comment. Distilled AI models can raise security concerns since they lack the safeguards of the models they were trained on, Anthropic wrote in a February blog post. These distillation campaigns allow actors to deliberately strip security protocols from the resulting models and undo mechanisms that ensure those AI models are ideologically neutral and truth-seeking, according to the director.
CNN has reached out to the Chinese embassy in Washington for comment. The Trump administration plans to share information with US AI companies about attempts to conduct unauthorized, industrial-scale distillation. The Trump administration plans to improve coordination between private companies in response to the campaigns.
The Trump administration plans to work with the private sector to develop best practices for finding and defending against industrial-scale distillation activities. Tech giants like Nvidia have been caught in the middle of US-China AI trade tensions.
The government has previously accused China of targeting American AI technology and intellectual property. The US has accused China of industrial-scale theft of AI.
Cementing the US as a leader in the AI race has been a cornerstone of Trump’s second term.
Trump has pushed for federal AI regulation rather than state-level rules. Trump’s push for federal AI regulation aims to expedite innovation. Trump has exerted controls over sales of AI chips to China.
