AI overly affirms users asking for personal advice | Researchers found chatbots are overly agreeable when giving interpersonal advice, affirming users’ behavior even when harmful or illegal.