On July 22, Michael Kratsios, director of the White House Office of Science and Technology Policy (OSTP), suddenly claimed on social media that Chinese AI company Moonshot AI had developed its Kimi K3 model by distilling Fable, a model developed by U.S. AI company Anthropic. Just hours later, U.S. Treasury Secretary Scott Bessent threatened that Washington would use sanctions and "Entity List restrictions" to respond to what he described as Chinese companies' use of distillation.

The sequence is familiar: First throw out accusations of "intellectual property theft," then impose restrictions on Chinese entities. In recent years, this has become a recurring pattern in U.S. efforts to curb China's technological development — portraying itself as the victim to take measures against others.

Such rhetoric may appear convincing to those unfamiliar with the technical details. Yet this time, the narrative is likely to fall flat, because the claim that Kimi K3's capabilities were achieved through distilling a U.S. model reveals a fundamental misunderstanding of how modern AI systems are developed.

One obvious issue is that, just as copying homework won't earn you good grades, Kimi K3's advanced performance simply cannot be achieved through distillation alone. Nathan Lambert, a prominent U.S. AI technology expert, has pointed out that Kimi K3 ranks second globally on the Vals AI Index, third on the Artificial Analysis Intelligence Index, and first on the Frontend Code Arena.

Lambert stated, "AI observers who followed the distillation panic and came away with the wrong conclusion that Chinese AI labs are only producing good models due to IP theft are in for an awakening — that Chinese companies are extremely good at building models in the same way the leading American companies are."

Lambert has also praised the Kimi research team's engineering culture. After visiting China and meeting some of the company's core researchers, he wrote, "It was clear to me that they had incredible culture, some would say aura, and a freedom to express it — within the constraints of a GPU-limited environment."

Another key point is that Kimi K3 is an open-source large language model. Unlike closed-source models developed by companies such as OpenAI and Anthropic, whose internal technologies remain largely inaccessible to the outside world, Kimi K3's technical details have been publicly disclosed.

In its technical report, the Kimi team introduced several architectural innovations, including a hybrid linear attention and Attention Residuals, which reportedly improved overall scaling efficiency by approximately 2.5 times compared with the previous-generation Kimi K2 model. The model reportedly contains 2.8 trillion parameters.

Some U.S. politicians fail to understand these technical advances — or choose to ignore them — when accusing Kimi K3 of relying on the distillation of U.S. models for its development. But researchers around the world can distinguish between genuine technological innovation and simple imitation.

The success of Kimi K3 ultimately comes from technological breakthroughs. As Stephen Hsu, a professor of Computational Mathematics, Science, and Engineering at Michigan State University, wrote on social media, the Kimi model's breakthrough "lies in fundamentally re-architecting the model's computational paradigm through innovations."

China's progress and achievements in AI have become increasingly visible to the global community. For U.S. policymakers holding influential positions in technology and national security debates, a deeper understanding of AI development would be preferable to making technically inaccurate accusations.

Attempts to dismiss China's advances as the result of "distillation" or imitation do not weaken China's AI capabilities. Instead, they make the accusers themselves appear disconnected from the realities of technological innovation.