Updated
Updated · The Atlantic · Sep 8
JAMA Commentary Urges AI-Only Care, Citing 15% Misdiagnosis Rate
Updated
Updated · The Atlantic · Sep 8

JAMA Commentary Urges AI-Only Care, Citing 15% Misdiagnosis Rate

3 articles · Updated · The Atlantic · Sep 8

Summary

  • A JAMA commentary argued autonomous medical AI should be tested as a lead clinician, saying AI-only care could soon outperform human doctors rather than remain a support tool.
  • Up to 15% of U.S. patients are misdiagnosed each year, the authors said, pointing to studies where specialized models beat physicians in diagnosis interviews, insulin dosing and perceived empathy.
  • A 2025 review of 52 studies found physician-AI hybrids generally did not beat AI alone or the best clinicians, though some stronger hybrid results came when AI made the final decision.
  • The American Medical Association rejected that conclusion, with CEO John Whyte criticizing the evidence base as too reliant on simulations and short-term studies, while the American College of Physicians said fully autonomous AI should stay limited to low-risk cases.
  • The clash widens a broader debate over whether tighter regulation and real-world trials should slow or accelerate a shift toward autonomous AI in healthcare.

Insights

If an autonomous medical AI makes a fatal misdiagnosis, who ultimately takes the blame in a court of law?
Could replacing rushed doctors with algorithms actually improve patient care, or are we trusting fragile tech with our lives?
When AI models fail unexpectedly in real hospitals, how can patients be sure their treatment plan is not based on a glitch?