Did Moonshot steal from Anthropic? US claims Kimi K3 was distilled from Claude Fable

The US administration has accused Moonshot AI of using Anthropic's Fable model to develop Kimi K3 and is considering tighter restrictions on Chinese AI. But Nvidia, Elon Musk and hundreds of Silicon Valley companies say banning Chinese AI models would do more harm to America than China.

In Short

• US alleges Moonshot AI used Anthropic's technology to build Kimi K3
• US officials said Moonshot accessed GB300 servers in Thailand for training
• Nearly 200 Silicon Valley companies urge Trump not to ban Chinese AI models

The US is accusing Chinese AI startup Moonshot AI of intellectual property theft. The Trump administration claims the company's latest AI model, Kimi K3, was built using technology distilled from Anthropic's AI models. The allegations have reignited calls in Washington for tougher action against Chinese AI firms. But many of the biggest names in the US tech industry, including Nvidia CEO Jensen Huang and Elon Musk, are pushing back, saying America should not fear Chinese open-weight AI models.

Speaking on Wednesday, White House Office of Science and Technology Policy Director Michael Kratsios accused Moonshot AI of using Anthropic's Fable AI model to build Kimi K3. He said the administration had information suggesting the company had created a "sophisticated internal platform" to carry out large-scale distillation of US AI models while attempting to avoid detection.

"We have information that Moonshot AI distilled Anthropic's Fable for the development of its K3 model," Kratsios wrote in a post on X.

"To do this, they developed a sophisticated internal platform to conduct large-scale distillation against US models, allowing them to quickly switch between multiple methods of access to avoid detection."

Kratsios claimed Moonshot AI had acquired Nvidia GB300-equipped servers and accessed additional GB300 systems in Thailand to train its models. While he acknowledged that AI distillation is a common technique used to create smaller and more efficient models, he argued there was a clear difference between legitimate distillation and what he described as "large-scale, covert industrial distillation" aimed at stealing proprietary American technology.

Could the US ban Kimi K3?

Kratsios' remarks came shortly after US Treasury Secretary Scott Bessent said the administration was looking into whether Chinese AI companies had improperly distilled American AI models. He added that companies found to have stolen US intellectual property could face sanctions.

Even so, the US is not currently planning a blanket ban on Chinese AI models or companies. Instead, the Trump administration is considering targeted restrictions aimed at reducing American businesses' reliance on Chinese AI platforms without imposing sweeping federal controls.

Why is the US worried about Kimi K3?

At the centre of the controversy is Kimi K3, Moonshot AI's latest open-weight large language model. Unlike closed AI models from companies such as OpenAI and Anthropic, open-weight models let developers download, customise and run them on their own infrastructure.

As the model is getting popularity, US officials are questioning whether it was built using technology improperly distilled from American AI models, while also worrying that China's AI industry is rapidly closing the gap with the US.

US tech industry calls for no ban

Meanwhile, several major US tech companies and startup founders are urging the Trump administration not to restrict access to Chinese open-weight AI models.

In an interview with Axios, Nvidia CEO Jensen Huang said American companies should "absolutely" be allowed to use Chinese AI models and argued that banning them would do more harm than good.

Huang described Chinese open-source AI as "excellent" and dismissed concerns that downloading such models would automatically expose companies to security risks or hidden backdoors. Instead, he said organisations can customise the models and deploy them within secure environments. He also argued that open models improve security because researchers can inspect them, identify weaknesses and build stronger safeguards.

According to Huang, the market has misunderstood the impact of models like Kimi K3, much as it did with DeepSeek earlier this year. Rather than reducing demand for AI hardware, he believes cheaper and more accessible AI models will encourage wider adoption, driving higher demand for Nvidia's chips, data centres and computing infrastructure.

Tesla CEO Elon Musk also appeared to support Huang's position, responding "True" to a post highlighting the Nvidia chief's argument that the US should not fear Chinese AI models.

Silicon Valley says a ban would hurt US innovation

Meanwhile, nearly 200 Silicon Valley companies, including startup accelerator Y Combinator and encrypted email provider Proton, have urged the Trump administration not to restrict access to Chinese open-weight AI models.

In letters sent by the newly formed Little Tech Association, the companies argued that American startups rely on affordable open AI models, many of which cannot easily be replaced by more expensive proprietary alternatives from OpenAI or Anthropic, reports Politico. They warned that broad restrictions would hurt US innovation more than China's. Instead of a blanket ban, they called for targeted safeguards against intellectual property theft while allowing developers continued access to AI models already available worldwide.

- Ends