19h agoAI researcher Janus argues Anthropic's Claude misattributes speakers in conversations to mimic human social status dynamicsSafety researcher @davidad compared the behavior to academic disclaimers.SentimentSentimentPos50%Neg50%Positive users value Claude's good suggestions and helpful criticism despite attribution errors, while negative users dislike its weird strawman pushbacks and status-loaded speaker mixups.8 comments with sentiment. View comments.