Key Takeaways Artificial intelligence (AI) systems increasingly marketed for healthcare may perform well on medical knowledge tests yet produce recommendations that can cause serious patient harm, according to a new safety benchmark developed by researchers at Stanford University, Harvard Medical School, and other institutions. The study, available in preprint and thus not yet formally peer-reviewed, […]
Key Takeaways Artificial intelligence (AI) systems increasingly marketed for healthcare may perform well on medical knowledge tests yet produce recommendations that can cause serious patient harm, according to a new safety benchmark developed by researchers at Stanford University, Harvard Medical School, and other institutions. The study, available in preprint and thus not yet formally peer-reviewed, […]
