JAMA + AI Weekly Update February 28, 2026

Can we trust AI-driven decision tools to uphold medicine’s “do no harm” principle? In the latest JAMA+ AI Conversations, JAMA Associate Editor Yulin Hswen, ScD, interviews Mass General Brigham’s David Wu, MD, PhD, and Beth Israel Deaconess Medical Center’s Adam Rodman, MD, MPH, about their work evaluating the safety and reliability of AI in health care.

They describe their live leaderboard that tracks errors made by large language models, from frontier models to specialized medical tools. Their findings show that errors of omission are more common than errors of commission, echoing patterns seen in human clinicians.

As more physicians routinely adopt AI for clinical support, understanding how this impacts clinical decisions only becomes more critical. The episode contains practical lessons, chess analogies, discussion of medical education, and consideration of how physician-AI teams might work.

Listen now on Spotify | Apple Podcasts | YouTube | JAMA.com.

Editor’s Picks in this week’s JAMA+ AI:

• A study using 10 LLMs found that clinical recommendations for simulated gastroenterology cases varied by patient demographics. Marginalized groups—transgender, unhoused, low-income, unemployed—were more often recommended to have mental health assessment. The findings highlight the risk of perpetuating biases when deploying AI in clinical decision-making. (JAMA Network Open)
• Diabetic retinopathy screening rates have historically been low among underserved populations due to barriers in accessing traditional eye care. The Diabetic Retinopathy Screening Point-of-Care Artificial Intelligence trial aims to demonstrate that a multicomponent approach within federally qualified health centers can improve patient adherence to annual retinal screening and diabetes standard of care. (JAMA Network Open)
• A 2025 executive order seeks to preempt most state-level AI regulations, including those that relate to medical AI, by pushing federal agencies to challenge or override state laws. In a commentary, the authors argue that this order exceeds presidential authority and risks undermining the balance of federalism, stifling state-led innovations in health care, and creating regulatory gaps. (JAMA)