The U.S. government, through White House official Michael Kratsios, has accused Chinese AI startup Moonshot AI of stealing Anthropic's Fable model to develop its Kimi K3 AI model. Kratsios claims Moonshot used 'distillation'—training a smaller model on the outputs of a larger one—to create K3, which is a 2.8 trillion-parameter open-weight model that outperforms U.S. models like Anthropic's Fable 5 and GPT-5.6. The accusation includes allegations that Moonshot built an internal platform to hide its activities and acquired restricted Nvidia GB300 chips for training in Thailand. Treasury Secretary Scott Bessent suggested potential sanctions against Chinese companies over such alleged intellectual property theft. Moonshot has not responded, nor has the Chinese embassy in Washington. While distillation is a standard practice, Kratsios argues the scale and secrecy of Moonshot's actions threaten U.S. technological leadership and national security.
Bias read (Conservative): The article frames the U.S. government's accusations as a justified defense of American technological sovereignty, emphasizing national security concerns and the threat posed by Chinese AI firms. The language suggests a strong alignment with U.S. government positions, particularly regarding the need





