Moonshot’s Kimi rattled markets. U.S. agencies now say it was trained on American models

U.S. agencies have asserted that Moonshot’s Kimi model, which drew significant market attention earlier this year, was trained using American models. The claim ties the origins of a high-profile product back to U.S.-based systems, according to the agencies making the allegation.

Moonshot is not named in isolation: it is listed among six Chinese AI companies that U.S. authorities say engaged in systematic extraction of capabilities from U.S. models. The agencies’ characterization frames the issue as a pattern of behavior across multiple firms rather than an isolated instance.

The disclosure adds a new dimension to public discussion about Kimi, which previously became notable for its market impact. The agencies’ assertion focuses on the model training process and the sources of capabilities, rather than on downstream uses or specific harms.

At this stage the allegations come from U.S. authorities and identify a group of firms said to have drawn on American models. Further details about the investigations, technical evidence, or responses from the companies named have not been provided in the information released alongside the allegation.