Full Breakdown
U.S. Agencies Accuse Chinese Artificial Intelligence (AI) Firms of Large-Scale Distillation Theft Ahead of Trump-Xi Talks
9/10/2026, 2:05:31 AM
Core Accusation
A joint cybersecurity advisory issued by the Federal Bureau of Investigation, the National Security Agency, and the Cybersecurity and Infrastructure Security Agency alleges that six Chinese artificial-intelligence companies have conducted “industrial-scale knowledge distillation campaigns” that extract proprietary capabilities from frontier U.S. models. The companies named are DeepSeek, Moonshot AI, Alibaba, MiniMax, StepFun and Z.AI. The advisory states that the campaigns have harvested “billions of tokens” and generated “millions of exchanges” with models from OpenAI, Anthropic, Google and xAI, and that the activity has occurred “likely with Chinese government awareness” since at least late 2024.
Background & Context
Knowledge-distillation is a technique in which a smaller model is trained on the outputs of a larger, more capable system. While widely used for efficiency, U.S. officials contend that the Chinese campaigns go beyond legitimate research, constituting systematic theft of intellectual property. The accusations surface as the United States and China prepare for a high-level meeting later this month between President Donald Trump and President Xi Jinping, where artificial-intelligence safety and competition are expected to be key agenda items. Earlier, on May 15, President Trump met President Xi in Beijing, underscoring the diplomatic backdrop to the current dispute.
Data & Statistics
- Companies targeted: DeepSeek, Moonshot AI, Alibaba, MiniMax, StepFun, Z.AI.
- U.S. models cited: OpenAI’s ChatGPT, Anthropic’s Claude, Google’s Gemini, xAI’s Grok, and SpaceX’s AI offerings.
- Scale of activity: “Billions of tokens” distilled; “millions of exchanges and requests” recorded.
- Timeframe: Activity documented as ongoing since at least late 2024.
Official Statements & Responses
- The Ministry of Commerce in Beijing dismissed the claims as “groundless,” characterizing the U.S. approach as a “pretext to achieve industrial monopoly” and accusing Washington of “double standards.”
Verbatim Quotes
- “The Chinese distill our models and they can never get ahead of us,” — Scott Bessent, US treasury secretary
- “Legitimate AI distillation used to create smaller, more efficient models plays a vital role in this open innovation ecosystem,” — Michael Kratsios
Why It Matters
The dispute highlights competing narratives over what constitutes legitimate AI research versus intellectual-property theft. U.S. officials link the alleged campaigns to potential military and cyber-attack capabilities, suggesting national-security implications. Conversely, Chinese officials frame the issue as a standard industry practice and call for cooperative governance. The tension may shape forthcoming policy measures, including possible sanctions, export controls, or bilateral agreements on AI safety.
Conflicting Reports & Gaps
U.S. agencies assert that the distillation activities were conducted “likely with Chinese government awareness,” implying state involvement. Chinese ministries counter that the accusations are “groundless” and portray the U.S. stance as an attempt to monopolize the AI sector. No independent verification of government awareness has been presented, leaving the degree of state participation unresolved.
What’s Next
The two leaders are slated to meet later this month at the White House, with AI expected to be a prominent discussion point. Both sides have signaled a willingness to engage, but the advisory and China’s rebuttal suggest that the upcoming talks will need to address divergent views on AI development practices and related security concerns.
