Our research goal for AI co-clinician is to support medical decision making with high-quality evidence.
We tested the system while adapting the NOHARM safety framework, and found it made zero critical errors in 97 of 98 primary care queries - outperforming comparable systems in blind evaluations.