Anthropic Claims Moonshot AI Bypassed Kimi Requests to Claude

Anthropic has alleged that Moonshot AI redirected around 300,000 user requests intended for Kimi to Claude instead. In a report released on September 10, Anthropic stated that most of these requests were processed through Claude's high-performance model Opus during a 10-day period, and users may have still believed they were using Kimi. Huaxing Finance cited Anthropic's report in claiming Moonshot AI used 5,380违规 accounts in the process. Anthropic also claims that Moonshot AI extracted Claude's reasoning process through 'session replay' after storing some interactions for internal model training, with over 23 million such interactions occurring between May and July, including requests containing corporate code and valid authentication information.
Korean Source
This article is an English localization of a Korean-language crypto news report. Original headline: 앤스로픽 “문샷AI, Kimi 요청 30만건 클로드로 우회”