The White House has formally accused Chinese AI startup Moonshot AI of secretly using Anthropic’s Fable model to build its recently launched Kimi K3 system. Michael Kratsios, director of the White House Office of Science and Technology Policy, posted on X on July 22, 2026, that officials had evidence Moonshot AI developed a covert internal platform specifically for large-scale distillation of U.S. AI models—switching between access methods to avoid detection.
According to Kratsios, the platform allowed Moonshot AI to extract capabilities from American models and then quickly change access methods. He also alleged the company acquired servers in Thailand to expand its computing capacity for training. Kimi K3, which debuted on July 16, has been hailed as one of China’s most capable AI models.
The timeline has sparked debate among AI researchers. Anthropic’s Fable 5 was re‑released on July 1 after export restrictions were lifted, leaving just 15 days before Kimi K3’s launch. Elie Bakouch of Prime Intellect and Dean Ball from OpenAI both questioned whether distillation alone could explain K3’s performance in such a short period, saying the claim does not make technical sense. The White House has not released public evidence, and Moonshot AI has yet to respond.
U.S. Treasury Secretary Scott Bessent warned that large-scale covert distillation of American intellectual property could lead to sanctions and placement on the Entity List, which restricts access to U.S. technology. Kratsios drew a line between legitimate model compression and what he called “theft.” The dispute adds tension ahead of potential IPOs by both Moonshot AI and Anthropic, and the administration says its response is ongoing.
