Full Breakdown
U.S. Administration Targets Chinese AI Model Extraction Amid Rising Tensions
4/25/2026, 4:00:29 AM
Executive Memo Calls for Crackdown on AI Model Distillation
On Thursday, President Donald Trump’s chief science and technology adviser, Michael Kratos, issued a memo accusing foreign entities “principally based in China” of running “industrial-scale campaigns to ‘distill’ capabilities from leading AI systems made in the United States” and of “exploiting American expertise and innovation.” The memo directs the administration to partner with U.S. AI firms to identify such activities, develop defensive measures, and pursue punitive actions against offenders.
Background: AI Competition and Model Extraction Threats
The memo arrives as the gap between U.S. and Chinese top-tier AI models has “effectively closed,” according to Stanford’s Institute for Human-Centered AI. Chinese startup DeepSeek recently released a large-language model that rivals U.S. offerings at a fraction of the cost, prompting U.S. companies to allege that the model was built through unauthorized “distillation” of proprietary U.S. systems.
Key Figures and Stakeholders
- Michael Kratos – White House chief science and technology adviser, author of the anti-extraction memo.
- Bill Huizenga (R-MI) – Representative who sponsored the House bill targeting foreign model extraction.
- David Sacks – Former AI and crypto adviser to President Trump, who linked DeepSeek to OpenAI’s models.
- Anthropic – U.S. AI lab that accused Chinese labs of illicitly extracting its Claude model.
- Liu Pengyu – Spokesperson for the Chinese embassy in Washington.
- Guo Jiakun – Spokesperson for China’s Foreign Ministry.
- Kyle Chan – Brookings Institution fellow specializing in China’s technology development.
Legislative Action: House Bill on Foreign AI Model Extraction
The House Foreign Affairs Committee unanimously backed a bipartisan bill that would create a process to identify foreign actors extracting “key technical features” from closed-source U.S. AI models. Sanctions and other punitive measures are outlined as possible responses. The bill’s ultimate reach remains uncertain, and analysts note that President Trump may avoid antagonizing China ahead of a planned mid-May state visit by President Xi Jinping.
Industry Allegations and Technical Concerns
OpenAI and Anthropic have formally alleged that DeepSeek and other China-based labs used distillation to appropriate their models. Anthropic described the practice as “illicitly extracting Claude’s capabilities to improve their own models,” warning that such shortcuts allow competitors to acquire powerful AI functions “in a fraction of the time and cost” required for independent development. Brookings fellow Kyle Chan cautioned that distinguishing unauthorized distillation from legitimate data requests will be “like looking for needles in an enormous haystack,” but suggested that coordinated information sharing among U.S. labs could mitigate the threat.
Official Statements from U.S. and Chinese Authorities
Kratos framed the issue as a national-security priority, urging the government to protect U.S. AI leadership. Huizenga labeled model-extraction attacks “the latest frontier of Chinese economic coercion and theft of U.S. intellectual property.” The Chinese embassy’s Liu Pengyu emphasized China’s commitment to scientific cooperation and intellectual-property protection, while Foreign Ministry spokesperson Guo Jiakun condemned the U.S. claims as unfounded and called for factual, prejudice-free dialogue.
Criticism and Opposition
Experts warn that the proposed enforcement mechanisms may be difficult to implement without stifling legitimate research collaboration. Chan highlighted the challenge of policing a “vast volume of legitimate requests for data,” suggesting that overly broad sanctions could hinder global AI innovation. Industry observers also note the risk of escalating trade tensions that could affect broader technology exchanges.
Verbatim Quotes
- “Model extraction attacks are the latest frontier of Chinese economic coercion and theft of U.S. intellectual property,” — Bill Huizenga, Representative (R-MI)
- “There's substantial evidence that what DeepSeek did here is they distilled the knowledge out of OpenAI's models,” — David Sacks, former AI and crypto adviser to President Trump
- “illicitly extract Claude's capabilities to improve their own models” — Anthropic (AI lab)
- “China firmly opposes this. We urge the U.S. to respect facts, discard prejudice, stop suppressing China's technological development, and do more to promote scientific and technological exchange and cooperation between the two countries,” — Guo Jiakun, Foreign Ministry Spokesperson
Implications and What’s Next
The administration’s crackdown signals a shift toward treating AI model extraction as a security issue. Pending legislation could introduce sanctions against identified offenders, while U.S. AI firms are urged to enhance data-sharing protocols. The outcome of the upcoming U.S.–China state visit and the enforcement of the House bill will shape the trajectory of AI competition and cross-border collaboration in the months ahead.
