China rejects US allegations of large-scale AI model theft
Beijing has dismissed allegations by the FBI, NSA and CISA that Chinese AI companies used large-scale model distillation to benefit their systems.
China has rejected US allegations that DeepSeek, Alibaba and Moonshot AI carried out “industrial-scale” theft of American artificial intelligence models, escalating tensions over the technology ahead of a planned visit by President Xi Jinping to the United States.
The allegations were made in a joint advisory issued by the FBI, NSA and CISA. The agencies said the Chinese companies had distilled billions of tokens from US models since at least late 2024, possibly with the awareness of the Chinese government. The claims have not been independently verified in the source material.
China’s Commerce Ministry called the allegations unfounded and accused the US of seeking a monopoly over the AI industry. It warned that Beijing would take “resolute countermeasures” if Chinese AI companies were suppressed under the pretext of targeting distillation.
Distillation involves training a less capable model using outputs from a more advanced one. China said the method was commonplace across the industry, including among US companies. The US advisory argued that the alleged scale and sophistication meant it was central to the Chinese companies’ model development rather than a supplementary technique.
The Foreign Ministry urged Washington to refrain from “unfounded accusations or smears”. It said China’s AI development reflected technological self-reliance and called for cooperation on AI that was open, inclusive and beneficial.
The dispute comes weeks before Xi is expected to visit the US for a summit with President Donald Trump, with AI among the issues expected to be discussed. The Chinese companies named in the advisory had not immediately responded to Al Jazeera’s request for comment.
