White House Memo Puts China on Notice Over AI Model Exploitation
The Trump administration has formally pledged to crack down on what it describes as the deliberate exploitation of American artificial intelligence models by foreign technology companies, with China explicitly identified as the primary offender. In a memo issued Thursday, Michael Kratsios, the president's chief science and technology adviser, accused foreign entities "principally based in China" of conducting industrial-scale campaigns to "distill" — that is, extract capabilities from — leading AI systems developed in the United States, characterizing these efforts as "exploiting American expertise and innovation."
Kratsios stated that the administration would collaborate with American AI companies to identify such activities, construct defenses against them, and develop mechanisms to punish those responsible. The memo signals a significant escalation in the U.S. government's approach to AI intellectual property protection at a moment when China is rapidly closing the technological gap with American competitors.
The Narrowing AI Performance Gap
The timing of the memo is notable. A recent report from Stanford University's Institute for Human-Centered AI concluded that the U.S.-China gap in the performance of top AI models has "effectively closed," underscoring the urgency felt by the White House. The administration has consistently argued that American dominance in artificial intelligence is essential for setting global standards and securing both economic and military advantages.
China has pushed back firmly against these characterizations. Liu Pengyu, spokesperson for China's embassy in Washington, said Beijing "opposed the unjustified suppression of Chinese companies by the U.S." and emphasized that "China has always been committed to promoting scientific and technological progress through cooperation and healthy competition," adding that the country "attaches great importance to the protection of intellectual property rights."
In Beijing, Foreign Ministry spokesperson Guo Jiakun told reporters on Friday that the U.S. claims were groundless and amounted to smearing the achievements of China's AI industry. "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," he said.
Bipartisan Congressional Support for Sanctions Legislation
Kratsios' memo arrived during the same week that the House Foreign Affairs Committee gave unanimous, bipartisan approval to a bill that would establish a formal process for identifying foreign actors who extract "key technical features" from closed-source, U.S.-owned AI models. The legislation would empower authorities to impose sanctions and other punitive measures against offenders.
Rep. Bill Huizenga (R-Mich.), the bill's sponsor, framed the issue in stark terms: "Model extraction attacks are the latest frontier of Chinese economic coercion and theft of U.S. intellectual property. American AI models are demonstrating transformative cyber capabilities, and it is critical we prevent China from stealing these technological advancements."
DeepSeek and the Distillation Controversy
The administration's focus on model distillation is rooted in concerns that gained widespread attention last year when Chinese startup DeepSeek shocked U.S. markets by releasing a large language model capable of competing with American AI giants at a fraction of the development cost. David Sacks, who was serving at the time as President Donald Trump's AI and crypto adviser, publicly alleged that DeepSeek had copied U.S. models.
"There's substantial evidence that what DeepSeek did here is they distilled the knowledge out of OpenAI's models," Sacks stated.
OpenAI, the developer of ChatGPT, echoed those concerns in a February letter to U.S. lawmakers, arguing that China should not be permitted to advance "autocratic AI" by "appropriating and repackaging American innovation."
Similarly, Anthropic, the maker of the Claude chatbot, alleged in February that DeepSeek and two other China-based AI laboratories had engaged in campaigns to "illicitly extract Claude's capabilities to improve their own models" using the distillation technique. Anthropic described distillation as a process that "involves training a less capable model on the outputs of a stronger one," noting that while the method can be a legitimate training approach, it becomes problematic when competitors "use it to acquire powerful capabilities from other labs in a fraction of the time, and at a fraction of the cost, that it would take to develop them independently."
The Complexity of Enforcement
The debate is not entirely one-sided. San Francisco-based startup Anysphere, maker of the widely used coding tool Cursor, recently acknowledged that its latest product was built on an open-source model developed by Chinese company Moonshot AI, the maker of the chatbot Kimi. This illustrates the entangled nature of global AI development and the difficulty of drawing clear lines around legitimate versus illicit technology transfer.
Kyle Chan, a fellow at the Washington-based think tank The Brookings Institution and an expert on China's technology development, warned that enforcement would be exceptionally difficult. He described the challenge as "looking for needles in an enormous haystack," given the need to separate unauthorized distillation from the vast volume of legitimate AI model queries processed daily.
Chan nonetheless identified potential paths forward. Information sharing and coordinated monitoring among U.S. AI laboratories could help detect illicit distillation campaigns, and the federal government could play a meaningful facilitative role across the industry. However, Chan also cautioned that the legislative prospects for the House bill remain uncertain, noting that President Trump may be reluctant to escalate tensions with Chinese President Xi Jinping ahead of a planned mid-May state visit to Beijing.
What Model Distillation Actually Means
Understanding the technical dimension of this controversy is key to assessing its significance. Model distillation, in its legitimate form, is a widely accepted machine learning technique in which a smaller, more efficient model is trained using the outputs generated by a larger, more powerful model. This allows developers to create compact systems that retain much of the capability of their larger counterparts without the associated computational expense.
The controversy arises when entities systematically query a competitor's proprietary model at scale and use those outputs to train their own systems — effectively transferring capability without paying the research and development costs incurred by the original developer. It is this large-scale, allegedly unauthorized use of American AI outputs that U.S. officials and companies are now seeking to define, detect, and penalize.
- Distillation: Training a smaller model on the outputs of a larger, more capable one
- Alleged abuse: Systematically querying proprietary U.S. AI models to reproduce their capabilities at low cost
- Affected companies: OpenAI (ChatGPT), Anthropic (Claude), and other leading U.S. AI developers
- Proposed remedies: Coordinated industry defenses, federal facilitation, sanctions via proposed House legislation
As Washington and Beijing continue to spar over technological supremacy, the question of how to enforce intellectual property norms in the rapidly evolving AI landscape remains both technically complex and geopolitically sensitive.
Source: SecurityWeek