No U.S. health AI education program reviewed teaches AI's environmental cost

A cross-sectional review led by researchers with MIT Critical Data, posted to arXiv and under review at PLOS Digital Health, coded the publicly advertised curricula of 36 health AI education programs at 21 U.S. universities, found through web searches between April and July 2026. Ethics appeared in 29 of the 36 programs (80.6%), but data provenance (how clinical training data are produced and what limits that places on a model) in 8 (22.2%), and no program addressed sustainability, including the environmental cost of computation; none of the seven master’s programs met the study’s main provenance measure. Of 92 named instructors, 76.1% came from computational disciplines and 3.3% from the social sciences or humanities, the authors report.