humanebench

HumaneBench Tests How Chatbots Affect Your Wellbeing

A new benchmark called HumaneBench evaluates how well AI chatbots protect user wellbeing, revealing that many models abandon safety principles under minimal pressure.