US claims Chinese AI firms are carrying out ‘industrial-scale’ theft of trade secrets
Washington Accuses Beijing of Systematic Extraction of American AI Capabilities
Qwenews.com – Three federal agencies — the FBI, the National Security Agency, and the Cybersecurity and Infrastructure Security Agency — issued a joint advisory on Tuesday alleging that major Chinese artificial intelligence companies have been conducting what they described as "industrial-scale" campaigns to strip proprietary capabilities from American frontier models. The accusation lands at a moment of acute friction between the two powers over who controls the next generation of computing intelligence, and it arrives weeks before President Donald Trump is scheduled to host Chinese leader Xi Jinping in Washington, DC, where AI is expected to feature prominently on the agenda.
What Distillation Means and Why It Matters
At its core, the technique at issue is called distillation: a process in which a smaller, less capable model is trained by observing the outputs of a larger, more advanced one. In academic settings, distillation is a well-established method for compressing knowledge and reducing inference costs. American officials and industry leaders have long acknowledged its legitimacy as a research tool. What they object to, however, is the scale, targeting, and opacity of the campaigns directed at commercial American systems. The advisory argues that when distillation is deployed at massive volume against specific competitors' models — often through intermediaries designed to obscure the true origin of queries — it crosses a line from fair research into appropriation of trade secrets.
"The sheer scale of these campaigns and their sophistication indicate that distillation is not a supplement to these companies' AI model development, but the critical core of it," the advisory states.
The agencies explained that Chinese firms exploit what they called a "gray market of proxies known as 'transfer stations'" to circumvent geographic restrictions, usage terms, and technical safeguards embedded in American AI products. By routing requests through third-party intermediaries, the firms allegedly mask their identity and volume, making it difficult for the target companies to detect or contest the extraction.
Named Targets and Specific Allegations
The advisory singles out several Chinese technology companies, including DeepSeek and Alibaba Group, as having carried out "high-volume knowledge distillation campaigns" against American AI firms. The implication is that these firms are not merely experimenting with distillation in a lab; they are building their competitive product lines substantially on capabilities extracted from rivals' proprietary systems.
Anthropic, the developer of the Claude model family, provided a concrete illustration. The company alleged that three Chinese AI labs had orchestrated an "industrial-scale campaign" to pull capabilities out of Claude. According to Anthropic's own figures, those labs generated more than 16 million exchanges with the model through roughly 24,000 fraudulent accounts — a volume far beyond what any legitimate research partnership would require.
Sanctions Threat and Diplomatic Stakes
Treasury Secretary Scott Bessent escalated the matter into trade-policy territory, warning that punitive measures are ready if the behavior continues. Writing on X, he stated:
"When [People's Republic of China] firms conduct covert, industrial-scale distillation attacks that cross the line into IP theft, sanctions and Entity List designations will be on the table."
The Entity List, maintained by the Department of Commerce, restricts the flow of American technology and components to designated foreign entities. A designation there can effectively cut a company off from U.S. chips, software licenses, and investment channels — a consequence that would reshape the competitive landscape for any firm named.
The timing sharpens the stakes. Trump has indicated that artificial intelligence will be among the topics he raises when Xi visits Washington later this month. An advisory of this specificity, issued days before a summit, signals that the administration intends to enter the conversation with documented grievances rather than general complaints about intellectual property.
A Broader Pattern of Targeting
The distillation accusations do not exist in isolation. Google's Threat Intelligence Group released a report the same day noting that hackers of varied backgrounds are "increasingly targeting AI assets," including proprietary model weights, source code, and cloud compute allocations. The report framed the shift in blunt terms:
"This shift underscores that enterprise AI assets—from model weights to cloud compute quotas—are high-value targets for espionage, extortion, and resource theft."
Read together, the two documents paint a picture in which American AI infrastructure is under sustained pressure from multiple directions: state-adjacent firms extracting capabilities through structured campaigns, and opportunistic actors probing for vulnerabilities in model repositories and training pipelines.
Why Open-Weight Models Are Central to the Dispute
Chinese open-weight AI models have gained significant traction globally because they are highly customizable, can run entirely on-premises without cloud dependency, and typically carry lower inference costs than closed American systems. If those models were, as the advisory alleges, substantially informed by distilled American capabilities, the competitive advantage they offer may rest partly on appropriated rather than independently developed knowledge. That distinction matters not only for trade negotiations but for the intellectual-property frameworks that govern cross-border technology transfer.
Washington has requested comment from the Chinese Embassy, DeepSeek, and Alibaba regarding the advisory's findings. As of publication, no formal response had been issued. The question now facing both capitals is whether distillation at this scale will be treated as a trade dispute to be negotiated, an IP violation to be litigated, or a national-security matter to be sanctioned — and whether the upcoming Trump-Xi meeting will produce a framework for resolving it, or simply codify the split.
Related Reading
Frequently Asked Questions
What is US claims Chinese AI firms are carrying?US claims Chinese AI firms are carrying is the main topic of this guide. The article explains the context, practical details, and next steps readers should understand.
Why does US claims Chinese AI firms are carrying matter?US claims Chinese AI firms are carrying matters because readers are looking for a useful answer, not just a short summary. Good content should match search intent and help them decide what to do next.