A study found ChatGPT Health under-triaged over half of emergency medical cases tested The AI performed well in moderate cases but missed 52% of gold-standard emergencies Mental health crisis alerts triggered inconsistently in suicidal ideation scenarios
