Prepared end to end by AI. The editorial is Dr. Hemant L. Patel. This is a newspaper-style weekly, not a numbered insight series. The PDF is the issue.
A chatbot now reads the records of millions who ask it for medical advice. The first patient-facing language model won clearance. The FDA’s comment window closes 19 October.
In this issue
- FDA comments on its first generative-AI device framework close 19 October; a same-day petition asks whether these tools are the agency’s business at all.
- ChatGPT Health now reads provider and Apple Health records; the company says 230 million people ask health questions weekly.
- SimBioSys TumorSight Risk: first AI breast-MRI recurrence score cleared on a 2,129-patient retrospective validation — figures company-reported.
- Stanford MediSyn: open synthetic medical images matched task-specific models across six specialties and ten modalities.
- The AMA set a four-question test for clinical AI agents: what it does, what evidence it works, what it may do, and who is accountable.
- India: AI MedTech still runs on CDSCO guidance, not a dedicated statute.
“A clearance is a starting line. A discussion paper is a map of questions. Neither is a finish line.”
AI Health Weekly · Vol. I · No. 2 · 5 pages · Dr. Hemant L. Patel
If the viewer is blank on this device, download the PDF.
The week the evidence asked harder questions
A diagnostic model beat doctors on paper, a meta-analysis showed humans rarely correct a wrong machine, and the FDA’s first generative-AI docket closed under a pile of letters.
Read No. 3 →The week AI in health got audited
Not launches. Checks.
Read No. 1 →The week the evidence caught up with the cheques
Big long-horizon commitments landed in the same week as the strongest outcome data the sector has published all year.
Read Impact Weekly →Authored data series live on Insights.