OpenAI's o1 correctly diagnosed 67% of ER patients vs. 50-55% by triage doctors
67% vs. 55%. OpenAI's o1 just outperformed human triage doctors in a Harvard ER trial—raising hard questions about AI deployment in healthcare.
Why it matters
A peer-reviewed Harvard study showing AI diagnostic superiority in high-stakes medical settings signals both opportunity and regulatory friction for enterprise AI deployment. This is the kind of capability claim that will drive board conversations about liability, validation, and real-world AI trustworthiness.
The key facts
10 to knowOpenAI o1 diagnosed 67% of ER cases correctly
Triage doctors diagnosed 50-55% correctly
Harvard-led trial (peer-reviewed implied by Guardian coverage)
Emergency department setting (high-stakes, regulated domain)
Published May 2026 (recent)
OpenAI o1: 67% diagnostic accuracy on ER triage cases
Human triage doctors: 50-55% diagnostic accuracy
Source: Harvard trial
Published: April 30, 2026
HIGH_ENGAGEMENT: 179 HN points, 127 comments
Go to the source
Hacker Newstheguardian.com
Publisher excerpt: Article URL: Comments URL: Points: 179 # Comments: 127