US accuses China's Moonshot AI of distilling Anthropic Fable to build Kimi K3

The US has accused Chinese AI startup Moonshot of stealing and secretly copying Anthropic’s Claude Fable 5 to build its much-hyphed model Kimi K3.

The claims came from the White House AI adviser Michael Kratsios who posted on X arguing that Moonshot developed a special platform for large-scale distillation of the US models.

Taking to X, he wrote, “To do this they 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.”

“Moonshot AI has also acquired GB300-equipped servers and has accessed GB300s in Thailand, likely to train its AI models,” Kratsios added.

The chips in question are Nvidia’s top AI chips and China is not allowed to obtain them under the US ruling that blocked their sale.

At the heart of this controversy lies the distillation technique. This method revolves around training a new AI model on the answers generated by other advanced AI models. Though, distillation is legal if it is performed openly.

However, “large-scale, covert industrial distillation aimed at stealing proprietary U.S. technology and undermining American research is unacceptable.”

This is not the first time China has been accused of stealing the US tech for training and building its own AI models. In February, Anthropic reported that Moonshot used hundreds of fake accounts, some matching the public profiles of senior Moonshot staff, to route over 3.4 million Claude exchanges.

According to the US tech giant, the campaign was designed to target and test Claude's coding, reasoning, and vision capabilities.

The release of Moonshot’s Kimi K3 has sent shockwaves across Silicon Valley because of its unprecedented capabilities. Being equipped with 2.8 trillion parameters, it is the biggest open-weight model ever released.

The model is expected to go public by July 27, 2026