Google DeepMind Launches AI Co-Clinician Research Initiative
Google DeepMind unveiled AI co-clinician, a multimodal agent system for healthcare support tested with Harvard and Stanford Medicine.
Google DeepMind has unveiled AI co-clinician, a research initiative exploring how multimodal agents could support healthcare workers and patients. The system uses live video and audio to process physical symptoms in real-time — analyzing gait, breathing, and visible skin conditions. Testing conducted alongside researchers from Harvard Medical School and Stanford Medicine found the system matched or outperformed physicians in 68 of 140 assessed areas, including triage. Using an adapted NOHARM safety framework, it made zero critical errors in 97 of 98 primary care queries.
The architecture relies on a dual-agent design: a "Planner" agent continuously monitors a "Talker" agent to keep responses within safe clinical boundaries. Human physicians still outperformed the system at spotting critical red flags and guiding physical examinations — a distinction the initiative frames as augmentation rather than replacement. Google DeepMind is gradually expanding a clinician-facing trusted tester program to additional sites globally as research progresses.