Google DeepMind Launches AI Co-Clinician Research Program: Multimodal AI Agents Assist Doctors and Patients

AIMPACT News, May 1 (UTC+8), DeepMind announced the launch of the AI Co-Clinician research program on April 30, 2026, exploring how multimodal AI agents can support medical decision-making under physician supervision. The system interacts in real-time via video and audio, capable of collecting medical histories, guiding physical examinations, performing preliminary diagnostic reasoning (such as rotator cuff injuries or tendinitis), providing care recommendations, and clearly stating that it is not a doctor; all suggestions must be used under physician supervision.
Technical features: utilizing real-time multimodal processing (analyzing gait, respiration, rashes, etc.), a dual-agent architecture (Planner continuously monitors Talker to ensure safety boundaries), and the NOHARM safety framework.
Test results: among 98 primary care queries, 97 achieved zero critical errors; in 20 simulated scenarios and 140 evaluation metrics, 68 metrics matched or exceeded doctors’ performance (especially in triage), with human doctors still having a clear advantage in identifying key “red flag” signals and guiding examinations.
The project collaborates with Harvard Medical School, Stanford Medical School, and others, currently in the research stage, with clinical testing gradually expanding. DeepMind emphasizes that AI aims to serve as a “collaborative member” to enhance rather than replace clinical judgment. (Source: BlockBeats)