The United States government is significantly escalating its efforts to prevent Chinese technology companies from accessing and replicating the capabilities of advanced American artificial intelligence models. In a coordinated push that spans both the executive and legislative branches, Washington is targeting what it describes as “industrial-scale campaigns” by foreign entities to distill the knowledge embedded in proprietary US AI systems. This aggressive new stance marks a shift from controlling the export of physical hardware, such as advanced semiconductors, to attempting to lock down the intangible outputs and behaviors of the models themselves.
The latest escalation was formalized in a memo issued on Friday by Michael Kratsios, the director of the White House Office of Science and Technology Policy (OSTP). The memo explicitly accused foreign entities, “principally based in China,” of deliberately extracting capabilities from leading US AI systems. According to the Associated Press, the administration has pledged to work closely with domestic AI companies to identify these extraction activities, build stronger technical defenses, and develop mechanisms to punish the offending parties.
The Challenge of Model Distillation
The core issue driving this new policy push is a technique known as model distillation. In this process, a smaller, less capable AI model is trained using the outputs generated by a larger, more advanced model. By repeatedly querying the advanced model and learning from its responses, the smaller model can rapidly acquire sophisticated capabilities without the massive computational expense and time required to train a frontier model from scratch. This technique effectively allows competitors to bypass the immense research and development costs borne by the original creators.
The urgency of the US response is underscored by recent assessments of the global AI landscape. A report from Stanford University’s Institute for Human-Centred AI recently concluded that the performance gap between top US and Chinese AI models has “effectively closed.”
(Related: Stanford AI Index 2026: China has nearly erased America’s AI lead)
This rapid convergence has alarmed US policymakers, who view AI supremacy as a critical component of national security and economic competitiveness.
The allegations of distillation are not merely theoretical. In February, OpenAI sent a letter to lawmakers arguing that China should not be allowed to advance “autocratic AI” by “appropriating and repackaging American innovation.” Similarly, Anthropic accused DeepSeek and two other China-based AI labs of illicitly extracting Claude’s capabilities. Even Anysphere, the maker of the popular coding assistant Cursor, recently acknowledged that its latest product is based on an open-source model developed by the Chinese company Moonshot AI, highlighting the complex and interconnected nature of global AI development.
Legislative Action and Diplomatic Friction
The executive branch’s concerns are being mirrored in Congress. The House Foreign Affairs Committee recently demonstrated unanimous bipartisan support for a bill designed to identify foreign actors extracting “key technical features” of closed-source US AI models and punish them with sanctions. Representative Bill Huizenga, a Republican from Michigan and the bill’s sponsor, framed the issue in stark terms: “Model extraction attacks are the latest frontier of Chinese economic coercion and theft of US intellectual property.”
The Chinese government has strongly rejected these accusations. Liu Pengyu, a spokesperson for the Chinese embassy in Washington, stated that China “has always been committed to promoting scientific and technological progress through cooperation and healthy competition,” and emphasized the country’s commitment to intellectual property protection. On Saturday, Chinese Foreign Ministry spokesperson Guo Jiakun went further, dismissing the US claims as “groundless” and accusing Washington of “smearing the achievements of China’s artificial intelligence industry.” Guo urged the US to “respect facts, discard prejudice, stop suppressing China’s technological development.”
Enforcing these new restrictions will be a monumental technical and diplomatic challenge. Kyle Chan, a fellow at the Brookings Institution, noted that separating unauthorized distillation queries from legitimate user requests will be like “looking for needles in an enormous haystack.” Furthermore, the timing of this crackdown is politically sensitive. With a planned state visit by Chinese President Xi Jinping to Beijing in mid-May, the Trump administration may face pressure to balance its aggressive tech containment strategy with broader diplomatic objectives.
The outcome of this delicate balancing act will significantly shape the future trajectory of the global AI industry. If the US successfully implements these restrictions, it could force Chinese AI developers to rely entirely on domestic innovation, potentially slowing their progress in the short term but accelerating their drive for complete technological independence in the long run. Conversely, if the restrictions prove too difficult to enforce, the rapid convergence of US and Chinese AI capabilities will likely continue unabated, fundamentally altering the balance of technological power. The stakes are immense, and the battle lines are now being drawn not just around physical hardware, but around the very knowledge embedded within the algorithms themselves.