White House Accuses Chinese AI Startup of Stealing Anthropic Technology

The White House has accused Chinese AI startup Moonshot AI of stealing technology from US firm Anthropic to develop its open-source model Kimi K3. Trump's technology advisor, Michael Kratsios, posted on X that Moonshot AI 'distilled' Anthropic's Fable 5 model, a process where one AI system is trained using another's data. Kratsios alleged that Moonshot AI conducted a 'large-scale' operation to rapidly switch model access to avoid detection, and that the company used Nvidia chips. Anthropic's terms of service prohibit unauthorized scraping for distillation, and the company has called for legislative and regulatory action. The accusations follow Treasury Secretary Scott Bessent's comments on Fox Business, where he said the US government has found traces of American models in several Chinese AI systems and threatened sanctions. Anthropic's Fable 5 model, launched in early June, was temporarily suspended at US government request before reopening in July. Kimi K3, released on Thursday, is described as the most advanced open-source model, excelling in benchmarks, particularly in application programming. Some experts note that training an agent requires more than just distillation. The Trump administration has announced an investigation into the matter.

Global Impact

Economically, the threat of sanctions on Chinese AI models could disrupt the global AI market, which relies on cross-border collaboration and open-source innovation. Politically, this deepens US-China technological decoupling, with both sides imposing restrictions on AI and semiconductor trade.