Full Breakdown
White House Accuses Chinese Entities of Industrial-Scale AI Model Theft
4/25/2026, 8:20:54 AM
Allegations of Industrial-Scale Distillation
On April 23, 2026, Michael Kratsios, director of the White House Office of Science and Technology Policy, issued a memo stating that foreign actors “principally based in China” are conducting coordinated, industrial-scale campaigns to “distill” U.S. frontier AI systems, systematically extracting model capabilities that undermine American research and innovation.
Methodology and Scale
Distillation trains a smaller model on the outputs of a larger, proprietary system. Kratsios’ memo cites “tens of thousands of proxy accounts” and “jail-breaking” techniques to query target models repeatedly. Anthropic reported 24,000 fraudulent accounts generating 16 million exchanges with Claude; OpenAI and others have logged similar large-volume queries.
U.S. Policy and Legislative Response
The administration will share intelligence with AI firms, improve private-sector coordination, and explore sanctions or export controls against violators. Rep. Bill Huizenga (R-MI) introduced bipartisan legislation to sanction foreign actors engaged in “model extraction attacks,” authorizing punitive measures.
Industry Evidence and Stakeholder Positions
OpenAI, Anthropic, and Google DeepMind have warned that Chinese labs—DeepSeek, Moonshot AI, and MiniMax—have attempted unauthorized distillation of their models. Anthropic called the campaigns “growing in intensity and sophistication,” and DeepSeek’s debut highlighted the commercial impact of copied technology.
Chinese Government Counter-Response
The Chinese embassy, via Liu Pengyu, said China “has always been committed to promoting scientific and technological progress through cooperation and healthy competition.” Foreign Ministry spokesperson Guo Jiakun called the U.S. claims “entirely baseless” and a “slanderous smear” against China’s AI achievements.
Strategic Implications
The dispute occurs amid a broader U.S.–China AI race, with security officials warning that stripped-down models could be used for surveillance or cyber attacks. It also overlaps with pending decisions on NVIDIA AI-chip sales to China and the May 14 Trump-Xi summit.
Conflicting Reports & Gaps
U.S. officials cite internal intelligence and industry-reported query volumes, yet no publicly verifiable technical evidence has been released. Chinese officials deny wrongdoing and provide no counter-data, creating an evidentiary gap that hampers independent assessment.
Verbatim Quotes
- “These coordinated campaigns systemically extract capabilities from American AI models, exploiting American expertise and innovation,” — Michael Kratsios, OSTP Director
- “Model extraction attacks are the latest frontier of Chinese economic coercion and theft of U.S. intellectual property,” — Rep. Bill Huizenga, R-MI
- “These campaigns are growing in intensity and sophistication.” — Anthropic, AI security team
- “The US claims are entirely baseless. They are a slanderous smear against the achievements of China’s artificial intelligence industry. China firmly opposes this,” — Guo Jiakun, Foreign Ministry spokesperson
Upcoming Developments
The White House will finalize inter-agency guidance before the Trump-Xi summit, while Congress debates the Huizenga bill. Industry groups plan further joint defenses against unauthorized distillation, and both governments say the issue will shape upcoming trade and security negotiations.
