The U.S. administration has leveled serious accusations against the Chinese AI startup Moonshot AI. According to Michael Kratios, Director of the White House Office of Science and Technology Policy, the company used Anthropic's Fable model to create its latest development — Kimi K3. This is not about simple borrowing, but about systematic industrial distillation.

Kratios claims that Moonshot AI built a complex internal infrastructure for large-scale extraction of knowledge from American models. This platform allowed for rapid switching between different access methods to avoid detection. Additionally, the company purchased servers with GB300 chips and gained access to similar equipment in Thailand, which, according to the official, indicates targeted preparation for large-scale training.

It is important to emphasize the difference between legitimate distillation — a commonly accepted method of training smaller models on the responses of larger ones — and what Moonshot is accused of. Kratios called the covert industrial extraction of proprietary technologies unacceptable. U.S. Treasury Secretary Scott Bessent has already warned that if intellectual property theft is confirmed, sanctions will follow, and the company will be added to the export restriction list.

However, the accusations also have a flip side. Prime Intellect researcher Eli Bakush noted that only 15 days passed between the reopening of access to Fable 5 (July 1) and the release of Kimi K3 (July 16). He doubts that such a short window is sufficient to explain the high performance of the Chinese model solely through distillation. This view was supported by OpenAI's Head of Strategic Futures, Dean Ball, who stated that the quality of Kimi K3 cannot be explained by mere copying of responses.

A screenshot of a dialogue also surfaced on social media, where Kimi K3 allegedly identifies itself as Claude. However, the authenticity of this screenshot has not been confirmed, and the context of its acquisition is unknown. The Chinese Embassy in Washington called the accusations "pure slander."

This is not the first incident: in February, Anthropic already accused Moonshot AI, DeepSeek, and MiniMax of extracting Claude's capabilities through fake accounts. At that time, according to the company, Moonshot generated over 3.4 million interactions. The current incident is just another round of escalation in the U.S.-China technological confrontation.

My analysis: Technically, the claim that Kimi K3 is "just a distilled Fable" seems weak given the time frame. More likely, the U.S. side is using this case as a pretext to tighten export controls and sanctions policy. Their real value is not so much in proving theft as in creating a precedent to block Chinese companies' access to advanced chips and technologies. The market should be prepared for further fragmentation of the global AI industry.