The largest user study to date on large language models (LLMs) assisting the general public with medical decisions has found that these systems pose significant risks due to their tendency to provide inaccurate and inconsistent information. more