The US accused a leading Chinese AI startup of exploiting American tech to create its latest model, threatening sanctions in response to what Washington described as “industrial-scale” theft of intellectual property.
The White House alleged that Moonshot improperly used Nvidia chips and Anthropic’s Fable model to create Kimi K3, whose advanced capabilities caused a US tech selloff last week reminiscent of the DeepSeek shock.
Washington is stepping up efforts to deter “distillation,” whereby one model is trained on another’s outputs — a process it believes China is using to develop frontier AI that could threaten US national security. But you “can’t stop software at the border,” Semafor’s tech editor argued: “You prevent those risks by hardening your own defenses.”