AI Rivalry Escalates as U.S. Accuses Chinese Model Maker of Copying American Technology- A new controversy has emerged in the global artificial intelligence race after White House science advisor Michael Kratsios accused Chinese AI company Moonshot AI, the developer of the Kimi K3 open-weight language model, of using stolen American technology to build its system.
Kratsios alleged that Moonshot trained Kimi K3 by copying capabilities from Anthropic’s Fable language model and relied on advanced chips that are restricted from export to China. He described the alleged activity as “large-scale, covert industrial distillation” aimed at acquiring U.S. technology and weakening American research leadership.
The accusations come amid growing discussions in Washington over whether Chinese open-weight AI models should face new restrictions. However, Kratsios did not provide evidence or details about the sources behind his claims, and Moonshot has not publicly responded to questions about its model-training methods.
The allegations were echoed by U.S. Treasury Secretary Scott Bessent, who claimed that American AI models were being reflected in Chinese systems. Bessent said officials had found “watermarks” from U.S. large language models in Chinese models, though the government has not explained what those watermarks are or provided public evidence.
Experts Question the Distillation Claims
While concerns about AI technology transfer are increasing, several researchers have questioned whether model distillation alone could explain the advanced performance of Kimi K3.
AI distillation involves using a powerful “teacher” model to generate data that helps train another model. Developers may use large numbers of prompts and responses from an existing system to reproduce some of its capabilities. However, researchers say creating a highly capable model through this method requires significant time, data, and additional training techniques.
Braden Hancock, a researcher at the Laude Institute and co-founder of Snorkel AI, argued that the timeline makes a simple copying explanation unlikely. He noted that Fable had only recently become publicly available, leaving little time for a company to collect enough data, train a new model, and release a competitive system.
Nathan Lambert, an AI researcher at the Allen Institute for AI, also questioned the idea that distillation alone could allow rapid advances. He suggested that modern AI progress increasingly depends on advanced training methods, including reinforcement learning, rather than only copying outputs from existing models.
Why Distillation Matters in the AI Race
AI distillation has become a major issue in global technology competition because it raises questions about intellectual property, national security, and fair competition.
The process can be legitimate when used by researchers and companies to create smaller, more efficient models. However, governments are increasingly concerned that it could also be used to replicate expensive AI systems developed by leading companies.
The debate has intensified as China and the United States compete for leadership in artificial intelligence. Open-weight models, which allow developers to access and modify model parameters, have accelerated innovation but also raised concerns about how easily advanced AI capabilities can spread. Burger King Gives Customers a New Reason to Trust Their Whopper | MayaBurger King Gives Customers a New Reason to Trust Their Whopper | Maya
A Larger U.S.–China Technology Battle
The dispute over Kimi K3 reflects broader tensions between Washington and Beijing over advanced technologies, including semiconductors, artificial intelligence, and computing infrastructure.
The U.S. has already imposed export controls on certain advanced chips used for AI development, arguing that limiting access protects national security. China has responded by investing heavily in domestic AI research and semiconductor development.
As AI capabilities continue to advance, disputes over training data, model ownership, and technology transfer are likely to become increasingly central to the competition between the two countries.
For now, experts say more evidence is needed to determine whether Kimi K3’s performance resulted from improper copying, independent innovation, or a combination of multiple AI development techniques.
