Large language models frequently misrepresent verbal risk terms used in medicine, potentially amplifying patient misunderstandings and diverging from established clinical definitions, according to a ...