AI Diagnostic Tools Flood U.S. Hospitals With Little FDA Oversight

Executive Briefing

• Reveals ChatGPT outperformed hundreds of physicians in a Harvard-Stanford diagnostic study published in Science.
• Warns most generative-AI health tools bypass FDA review by classifying as 'clinical decision support' rather than medical devices.
• Documents errors including AI drafting 'completely absurd' patient messages and failing to flag potential medical emergencies.
• Notes consumer apps from Hims & Hers, Microsoft, and OpenAI collect medical records while disclaiming they provide no medical advice.
• Highlights a government-funded effort to put one cardiac AI chatbot through full FDA authorization as a potential regulatory model.