Mansi, Riedl: 31 US cases urge patient-first AI liability

Two researchers at Georgia Tech read 31 US legal cases involving AI in healthcare and concluded that pinning accountability on physicians misses where the harm actually happens.

The analysis in Nature Health, by Gennie Mansi and Mark Riedl and published on 21 September 2026, pulls cases from three databases (DAIL, the O'Neill Institute Health Litigation Tracker, and AIAAIC) reviewed as of February 2025. The authors argue that "patients' medical care relies on a complex web of stakeholders—physicians, state health departments, health insurers and care facilities, among others—and many AI tools deployed across their healthcare delivery negatively impact their care."

Their reading is blunter still: "patients have had no option but to seek legal recourse for harms."

From there, Mansi and Riedl lay out two design shifts. One, changing liability structures to reflect the range of actors who shape how an AI tool touches a patient. Two, building tools that help patient-side lawyers do the legal work that recourse actually requires.

The piece is framed as an analysis, not an empirical harms tally. The abstract publishes no per-case outcomes, no dollar figures, and no breakdown of which tools appear most.