Researchllmhealthcareopenaisafety evaluation
Researchers Find ChatGPT Health Misses Emergency Alerts
9.2
Researchers at the Icahn School of Medicine at Mount Sinai published a Nature Medicine study this week testing OpenAI’s ChatGPT Health, using 60 clinical scenarios and 960 interactions. They found the tool failed to recommend emergency care in 52% of serious cases and inconsistently flagged suicide-crisis risk. The authors urge independent evaluation, ongoing monitoring, and engineering safeguards for consumer-facing health AI.
Key Points
- 1Detected failures: ChatGPT Health missed emergency care recommendations in 52% of serious cases (960 interactions).
- 2Showed inconsistent suicide-crisis alerts, appearing more for lower-risk scenarios than high-risk disclosures.
- 3Implies need for independent evaluation, monitoring, and safeguards before deploying consumer-facing health AI.
Scoring Rationale
Strong peer-reviewed evidence of systemic safety failures, tempered by single-tool focus and need for broader replication.
Sources
Public references used for this report.
Practice with real Health & Insurance data
90 SQL & Python problems · 15 industry datasets
Used by DS/ML engineers at top companies
250 free problems · No credit card
See all Health & Insurance problems
