China India Japan Korea Southeast Asia Economy Politics
Home China Feature
China · Exclusive

US advisory urges AI firms to poison Chinese model distillation efforts

US advisory urges AI firms to poison Chinese model distillation efforts
China · 2026
Photo · Mei-Ling Chen for Asian Examiner
By Mei-Ling Chen China Correspondent Sep 11, 2026 3 min read

Washington has escalated its technological rivalry with Beijing by quietly urging American AI companies to sabotage Chinese efforts to copy their models, a move that has drawn sharp warnings from China and set the stage for contentious bilateral talks.

The advisory, issued jointly by the FBI, NSA, and CISA, accuses six Chinese AI firms—DeepSeek, Moonshot AI, Alibaba, MiniMax, StepFun, and Z.AI—of using industrial-scale knowledge distillation to extract billions of tokens from leading US models like Claude, ChatGPT, Google Gemini, and Grok since late 2024. The agencies allege these campaigns were conducted with the knowledge of the Chinese government.

Knowledge distillation is a common technique where a smaller model learns to mimic a larger one, allowing developers to approximate advanced capabilities at lower cost. The US advisory, however, frames it as a systematic theft of intellectual property, calling it “the core” of China's AI strategy.

Beyond recommending account blocking and monitoring, the advisory contains a less-publicized section that urges US firms to deploy “targeted changes” in responses to suspected distillation requests. These include using “downgraded” models, reducing reasoning depth, or introducing stylistic inconsistencies—essentially feeding false or misleading data to poison the training of Chinese models. The advisory explicitly advises against informing suspected distillers of such switches, to prevent them from adapting.

Beijing has reacted forcefully. Foreign Ministry spokesperson Mao Ning defended China's AI progress as self-reliant innovation and accused Washington of using national security as a pretext to protect its dominance. An unnamed Commerce Ministry official called the US approach a “textbook case” of using distillation accusations to justify industrial monopoly, noting that American firms have also drawn on Chinese open-source models.

The dispute comes weeks before scheduled AI safety talks in Beijing, where US Treasury Secretary Scott Bessent is expected to lead the American delegation. Bessent has framed the rivalry in stark terms, saying at a Breitbart forum, “There is no day after tomorrow if China wins at this.”

Chinese commentators argue that distillation is a vital shortcut for Chinese firms to overcome US export controls on advanced chips, which have been in place since late 2022. They see it as a legitimate method to catch up in a field where Washington seeks to maintain its edge.

The advisory's call for poisoning tactics has drawn little attention in global media, but it marks a significant escalation in the US-China AI rivalry. As both sides prepare for talks, the question is whether they can find common ground or whether the poisoning strategy will further deepen the divide.

For more on the broader US-China tech competition, see this analysis of the leverage game and how China's embedded AI strategy is shifting competition.

More from this story

Next article · Don't miss

Chip industry's next leap: merging memory and logic

Semiconductors are finally moving into the third dimension, with memory and logic beginning to merge. This shift, driven by AI demand and new engineering, is redrawing the industry's competitive map.

Read the story →
Chip industry's next leap: merging memory and logic