White House Accuses Moonshot AI of Distilling Anthropic's Tech to Build Kimi K3

On July 16, 2026, Chinese AI lab Moonshot AI released Kimi K3, a 2.8-trillion-parameter open-weight language model. Two days later, K3 reached the top of the Frontend Code Arena, scoring 1679 points and beating Anthropic's Claude Fable 5 and OpenAI's GPT-5.6 Sol on coding tests. The model was designed for long-horizon programming, knowledge work, and multimodal reasoning, with full model weights scheduled for public release on July 27. The result quickly became a political story. On July 22, the White House and the US Treasury accused Moonshot of distilling Anthropic's Claude Fable to train K3, claiming that Chinese labs were running large-scale AI distillation campaigns that copied US frontier models. US officials warned that any company caught doing this could face export controls, sanctions, and other restrictions. The allegation turned a coding benchmark into a test of US-China technology policy. Moonshot strongly denied the allegation. Huang Zhenxin, head of the company's enterprise business, said K3's leap in performance came from original architecture innovation, not distillation, and pushed back against the idea that open Chinese models should be labelled cheap. Markets also reacted. NVIDIA's market value fell by about 111 billion dollars the day after K3 was released, and Anthropic rolled back a planned price change to keep users from switching to cheaper open alternatives. The dispute is now bigger than one company. US regulators are watching every major Chinese AI release more closely, while Chinese labs argue that open-weight models are forcing the whole industry to compete on cost and quality. The escalation shows that the line between model performance and geopolitics is getting thinner, and a single coding benchmark result can now move hundreds of billions of dollars on Wall Street.