Prof Gillie Gabay

Generative AI can reduce clinical documentation burdens and improve communication efficiency, but its use introduces risks involving inaccurate outputs, automation bias, privacy, liability, and patient trust. Safe deployment requires mandatory human verification, transparent patient disclosure, strong data protections, and continuous auditing to ensure efficiency gains do not compromise clinical quality or safety

Clinical AI can improve diagnostic speed, efficiency, and patient care, but it must remain subordinate to human clinical judgment. A robust Human-in-the-Loop framework requires physician accountability, patient transparency, data protection, continuous algorithmic auditing, and measures to prevent automation bias and clinician deskilling.

Generative AI performs strongly with structured clinical data but struggles to interpret the informal, culturally diverse language patients use to describe symptoms, creating risks for safety, efficiency, liability, and health equity. Successful deployment requires real-world validation, human oversight, continuous monitoring, and patient-centered interfaces designed around natural communication.