Preserving clinical judgment in the age of generative AI

Abstract As generative artificial intelligence (AI) moves toward large-scale healthcare deployment, a central challenge is how to harness its benefits without eroding clinical expertise or accountability. This commentary, written from clinician and technologist perspectives, argues that responsible AI adoption must rest on two foundations: preserving human accountability for medical decisions and integrating machine learning with structured, explainable reasoning. AI offers significant promise in diagnostics, personalized therapeutics, workflow automation, and burnout reduction. Data-driven discovery and biotechnology advances further expand therapeutic possibilities. However, risks include hallucinations, automation bias, and progressive deskilling. Emerging evidence suggests sustained AI assistance may diminish unaided diagnostic reasoning, underscoring the need for safeguards that maintain core clinical skills. From an implementation standpoint, success depends less on technical novelty than on integration within established care pathways, regulatory clarity, and demonstrable safety. Governments and large health systems are well positioned to scale validated solutions. The most sustainable path forward is hybrid “neuro-symbolic” systems that combine generative AI’s conversational capabilities with transparent, probabilistic clinical reasoning. Trustworthy healthcare AI must remain explainable, auditable, and decisively human-led.