Amazon Tests Health Agents
From MyClaw Newsletter · subscribed via aiste.ulozaite@gmail.com · original ↗ · unsubscribe
Amazon introduced PatientAgentBench, a clinician-vetted framework that creates synthetic patient records and multiturn conversations to test healthcare AI agents on safety, triage, workflows, and task completion. Across thousands of evaluations, even frontier models missed crisis resources, fabricated clinical details, and mishandled routine requests hiding serious risks.
Couldn’t fetch the full article — read it on the original site ↗.