US Accuses Chinese AI Firms of Industrial-Scale Tech Theft
Three US federal agencies alleged Tuesday that Chinese AI firms are conducting large-scale theft of American AI trade secrets through a technique called distillation. The FBI, NSA, and CISA said the campaigns are central — not supplemental — to how those firms develop their models.
- Anthropic claimed 3 Chinese AI labs generated over 16 million exchanges via roughly 24,000 fraudulent accounts.
- DeepSeek and Alibaba Group were named as conducting high-volume distillation campaigns against US firms.
- Chinese firms used a gray market of proxy networks to evade US companies' geographic safeguards, agencies said.
- Google's Threat Intelligence Group separately reported hackers are increasingly targeting AI model weights and source code.
Why it matters: Distillation — training a weaker AI model on a more advanced one — can be legitimate, but US officials say the scale and targeting of Chinese campaigns cross into IP theft. The advisory marks the most detailed public evidence the government has offered for accusations that AI executives have made for some time.
- Sanctions and Entity List designations are now explicitly on the table, according to Treasury Secretary Scott Bessent.
How 5 sources split on this story
Left coverage, 2 sources: The left frames this as a significant escalation in the US-China tech rivalry, emphasizing the breadth of the alleged campaigns and the potential for sanctions while noting that distillation itself is a legitimate technique.
What’s next: Trump is expected to raise AI with Xi Jinping at a Washington meeting later this month.
- CNN said it sought comment from DeepSeek, Alibaba, and the Chinese Embassy; none had responded at publication.
- What evidence, if any, will the US government make public to support the specific company-level allegations?
- How will China respond diplomatically, and will AI theft accusations derail the upcoming Trump-Xi talks?