Anthropic reports Moonshot AI redirected nearly 300,000 Kimi requests to Claude

Anthropic released an AI abuse report stating that Moonshot AI (月之暗面) redirected nearly 300,000 user requests from its Kimi service to Claude over a 10-day period, mostly to the Opus model, using 5,380 unauthorized accounts. Users were unaware their requests were being processed by a third party. Anthropic also alleged that Moonshot AI used "cross-session replay" to extract Claude's reasoning processes to train its own models, resulting in over 23 million distillation interactions between May and July.