Moonshot AI Criticized by the White House for Kimi K3

Moonshot AI Accused of Technological Theft by the White House

The White House has recently intensified its criticism of Moonshot AI, a Chinese company, accusing it of illegally using technologies developed by Anthropic to create its artificial intelligence model, Kimi K3. This accusation was made by Michael Kratsios, Director of the White House Office of Science and Technology, who stated that Moonshot AI had employed sophisticated methods to appropriate Anthropic's advancements.

Kratsios claimed that the White House possesses evidence demonstrating that Moonshot AI used a process called distillation to extract elements from Anthropic's Fable model in order to develop Kimi K3. According to him, Moonshot AI established a complex internal platform to conduct this large-scale distillation, allowing the company to quickly change methods to evade detection.

Kimi K3: An Impressive Open Source AI Model

Moonshot AI's Kimi K3 model is widely hailed as one of the most impressive open source AI models available to date. In performance tests, Kimi K3 matched or even surpassed the state-of-the-art models developed by American companies such as Anthropic and OpenAI. These companies have likely invested billions of dollars in developing their models, while Moonshot AI plans to make Kimi K3 available for free, allowing users to download and customize it.

Distillation: A Controversial Practice

Distillation, while controversial, is not inherently illegal. Kratsios emphasized that this process is commonly used by AI labs to train smaller, more efficient models from more powerful ones. However, he criticized Moonshot AI's large-scale and secretive use of distillation, labeling it an attempt to steal American proprietary technologies.

Kratsios stated, "Legitimate AI distillation is essential for open innovation, but industrial and secretive use to undermine American research is unacceptable."

Reactions and Political Implications

This statement from the White House comes shortly after Scott Bessent, Secretary of the Treasury, mentioned the possibility of sanctions against Chinese companies if their AI models are proven to be improperly trained through distillation. This situation is particularly notable as the Trump administration, typically at odds with Anthropic, seems to be aligned with the company this time. Indeed, the administration had previously imposed export controls to prevent Anthropic from releasing its Fable 5 model, and in February, the Pentagon decided to blacklist this giant AI.

In February, Anthropic had already publicly accused several companies, including Moonshot AI, of using distillation techniques in violation of its terms of service. These accusations come amid rising tensions between the United States and China regarding intellectual property in the field of AI.

A spokesperson for Anthropic did not immediately respond to a request for comment, but the implications of these accusations could be significant for trade and technological relations between the two countries.