Clinical AI Safety Isn’t Magic: Building Frameworks Health Systems Can Trust
Information
As clinical AI rapidly enters healthcare workflows, health systems are increasingly concerned not just about accuracy, but also safety, transparency, and long-term clinical impact. Clinical AI must be governed with the same rigor as patient care itself.
In this session, Jonathan H. Chen, MD, PhD (Stanford University), Joshua Geleris, MD (SmarterDx), and Scott Fleming, PhD (SmarterDx) explore how healthcare organizations can safely evaluate, adopt, and govern clinical AI as it moves from experimental tools to real-world clinical infrastructure.
Drawing on Stanford research and real-world deployment experience, the speakers examine the rapid progress of large language models in medical reasoning while highlighting why benchmark performance alone is insufficient for safe deployment. The session presents a practical framework for AI safety that treats AI as a dynamic production system, offering health system leaders a clear checklist for evaluating vendors and ensuring clinical AI is transparent, accountable, and safe for patient care.
