White House accuses Moonshot of distilling Anthropic's Fable for Kimi K3; tech leaders defend open distillation

The White House Office of Science and Technology Policy Director Michael Kratsios accused China-based Moonshot AI of covertly distilling Anthropic's frontier Fable model to build Kimi K3, a 2.8-trillion-parameter open-weight model released last week. Kratsios stated Moonshot 'developed a sophisticated internal platform to conduct large-scale distillation against U.S. models, allowing them to quickly switch between multiple methods of access to avoid detection.' He also alleged Moonshot acquired restricted Nvidia GB300 chips via servers in Thailand to train the model.

Kimi K3 performs competitively with Anthropic's Fable 5 and substantially outperforms several other leading U.S. models in selected benchmarks, which sparked the White House concern. Model distillation is a legitimate technique in which a smaller model learns from outputs of a larger one; however, Kratsios framed the scale and covert nature as IP theft. Treasury Secretary Scott Bessent echoed the accusation, saying the U.S. is 'finding watermarks of our U.S. large language models on many of the Chinese models' and hinted sanctions could be imposed.

In response, a coalition of 25+ tech companies including Nvidia, Microsoft, Meta, and Palantir issued a joint letter Friday urging policymakers to avoid 'premature restrictions' on open-weight models that would 'stifle competition or drive innovation overseas.' The letter defended distillation as 'a widely used technique for model improvement, evolution, and validation.' Independent experts questioned whether Fable distillation alone could account for Kimi K3's performance, noting that advanced reinforcement learning techniques require significantly more infrastructure than simple distillation. The debate reflects a fundamental tension: whether distillation is a legitimate open-innovation practice or an unacceptable form of IP theft.