A new paper describes how, despite widespread enthusiasm about artificial intelligence’s potential to revolutionize healthcare and the use of AI-powered tools on millions of patients already, no federal regulations require that AI-powered tools be evaluated for potential harm or benefit to patients.