Story perspectives

OpenAI o1 AI Diagnoses 67% Cases, Outpaces Doctors**

5/4/2026

Summary

OpenAI o1 Beats Doctors

OpenAI’s o1 model diagnosed 67% of 76 ER triage cases, beating physicians who scored 55% and 50%.
Using raw EMR data, the o1 model outperformed the 4o model at every point, while Adam Rodman warned of no accountability framework and patient desire for human guidance.
Arjun Manrai said the AI eclipsed prior models, while the authors plan trials with patients and will test AI on images and lab results.

Summary

ERs Adopt AI

Emergency rooms have begun using generative AI to draft notes, shifting documentation toward automation.
The Harvard study frames generative AI as expanding from labs into offices, classrooms and courtrooms.
Researchers warn AI suggestions may raise costs and expose patients to unnecessary procedures, while physician Adam Rodman says patients want human guidance for critical choices.
The lack of a formal accountability framework is prompting calls for regulatory guidelines on AI-driven diagnoses.

Summary

Israel Approves AI

Israel’s Health Ministry approved the AI system LIV, which matched psychiatrists 90 % of the time and identified high-risk conditions in 96 % of 385 patients.
David Reich, chief clinical officer of Mount Sinai, says the system must be embedded into clinical workflows rather than merely proven accurate.
The Harvard team’s parallel studies on medical-image models, which report rapid improvement, show the AI’s management reasoning scored a median 89 % versus doctors’ 34 %.

1 of 3

OpenAI o1 Beats Doctors

OpenAI’s o1 model diagnosed 67% of 76 ER triage cases, beating physicians who scored 55% and 50%.
Using raw EMR data, the o1 model outperformed the 4o model at every point, while Adam Rodman warned of no accountability framework and patient desire for human guidance.
Arjun Manrai said the AI eclipsed prior models, while the authors plan trials with patients and will test AI on images and lab results.

1 / 3