Paging the algorithm: applying the Best Available Human principle to graduate medical education

Artificial intelligence (AI) is transforming graduate medical education (GME), yet formal training in its responsible use remains limited. As AI capabilities expand, trainees increasingly adopt these tools informally and without structured oversight, raising urgent questions about how to integrate AI into physician training and teach its responsible use. This article applies Ethan Mollick's Best Available Human (BAH) standard as a principle to guide trainee use of AI. BAH permits AI engagement when its performance meets that of the best human expertise readily available, and offers a simple, flexible rule to help trainees decide when AI can responsibly augment learning and patient care. BAH is adapted for GME by pairing its performance and availability threshold with structured verification and faculty to review to reinforce responsible AI use. The authors apply the BAH principle across three domains central to GME: 1) clinical instruction, 2) diagnostic reasoning, and 3) health record composition. In each domain BAH conditions AI use on availability thresholds, demonstrated performance, and structured verification. Because the BAH principle would function best when trainees possess a foundational understanding of AI tools and their appropriate use, the authors argue that GME programs should formally incorporate established AI competencies into their curricula. These competencies align with the skills needed to master the BAH principle and use AI responsibly. Together they make disciplined, responsible AI use both teachable and feasible within contemporary GME training programs.