Concerns Raised Over Kimi K3's Development and Use of Banned Chips

Jul 23, 2026

AI Summary

Michael Kratsios, a White House science advisor, criticized the Chinese company Moonshot for allegedly copying Anthropic's Fable LLM and using banned chips for the Kimi K3 model. Experts express skepticism about the effectiveness of distillation in achieving Kimi K3's capabilities, suggesting that more advanced techniques may have been employed.

Michael Kratsios, a White House science advisor, accused Moonshot of copying Anthropic's Fable LLM and using chips not cleared for export to China for its Kimi K3 model.
Treasury Secretary Scott Bessent noted that watermarks from U.S. models are found in many Chinese models, raising concerns about technology theft.
Experts, including Braden Hancock and Nathan Lambert, doubt that distillation alone could account for Kimi K3's advanced capabilities, suggesting that more complex training methods may have been used.
Distillation involves systematically querying a model to extract data, but its effectiveness is diminishing as models become more sophisticated.
Moonshot has been accused of systematically distilling models from Anthropic, which claimed to have identified unusual usage patterns from the company.
The use of advanced Nvidia chips by Moonshot, which are banned for export to China, raises further concerns about compliance with export regulations.
Experts emphasize the technical expertise of Chinese AI teams and the need for better oversight of data centers that allow training on advanced hardware.
The U.S. Department of Commerce proposed know-your-customer rules for data centers in 2024, but progress on this initiative has stalled.

anthropicfablemodel trainingdistillationexpert analysis

Concerns Raised Over Kimi K3's Development and Use of Banned Chips

Related Stories

Grok 4.6 Achieves 61 on Artificial Analysis Intelligence Index, Enhancing SpaceXAI's Position

Exploring the Mathematical Capabilities of Large Language Models

Google's Gemini app reaches over 1 billion monthly active users

Anthropic AI model advances understanding of the Riemann hypothesis