When AI Harms Patients, Courtrooms Reveal Who Gets Blamed -- and Who

A new analysis in Nature Health, authored by Georgia Tech researchers Gennie Mansi and Mark Riedl, examines 31 U.S. legal cases involving medical AI and argues that courtroom accountability often focuses too narrowly on physicians. The study finds that patient harm is shaped by a broad network of stakeholders—doctors, state health departments, insurers, care facilities, and device developers—while legal frameworks still tend to treat accountability as if one clinician controlled every algorithmic decision. The researchers categorize harms by clinical tasks such as diagnosis, testing, treatment planning, triage, documentation and communication, and care coordination, and report that injuries cluster not only around diagnostic tools but also around administrative and coverage systems, including insurer algorithms and state screening tools.