ai
AI Tools Enter U.S. Hospitals Without FDA Approval
AI diagnostic tools outperformed physicians in one study released in April. Hospitals have deployed multiple generative AI products for clinicians even though none have received FDA clearance for medical use.
8thlight.comAudio version
Tap play to generate a narrated version.
Developing·Limited corroboration so far. This page will refresh as more sources emerge.
A study published in the journal Science found that ChatGPT outperformed hundreds of physicians at diagnosing written medical cases drawn from real patients. The research was led primarily by researchers from Harvard and Stanford. Lead author Adam Rodman said the results raised questions about how the findings might be applied in practice.
Hospitals have begun rolling out AI clinical reasoning tools to physicians. One medical center notified staff that an AI-powered tool was available for use in patient care.
Many AI products used by doctors avoid FDA oversight by qualifying as clinical decision support tools. These tools must rely on existing medical literature, avoid analyzing scans, and leave final decisions to physicians. Consumer health apps from Microsoft, OpenAI, Anthropic, and xAI state they do not provide medical advice.
Some products still offer interpretations of lab results and imaging studies.
A randomized trial published in NEJM AI showed that intentionally incorrect AI output can lead doctors off course. Separate studies found AI did not improve self-diagnosis by nonprofessionals and sometimes failed to flag medical emergencies. A hospital spokesperson said use of the tools is voluntary and any output requires physician approval.
Microsoft AI vice president Dominic King said its Copilot app supplies information to support conversations with clinicians. Hims & Hers chief medical officer Patrick Carroll said its Labs AI product does not diagnose or recommend treatment.
