Researchers point to risks in high-stakes applications as well as the potential to spread misinformation

Large language models often fail to distinguish between factual knowledge and personal belief, and are especially poor at recognizing when a belief is false.…