Anthropic accuses DeepSeek and Moonshot-linked actors of distilling its frontier models

1a3orn · x · 2026-11-21

A screenshot of Anthropic's post on "the national security implications of adversarial distillation" reveals: in February Anthropic briefed Congress on activity indicating DeepSeek was attempting to distill U.S. frontier models; in September it disrupted a campaign to extract protected reasoning from its models, attributing a core part to individuals associated with Moonshot AI (Kimi's developer); more disrupted campaigns will be disclosed soon. The poster quips the document "is not written for you or me."

Original post →

More from Companies & People

Companies & People channel →