A new study argues that certain frontier chatbots are much more likely to inappropriately validate users’ delusional ideas — ...