Nature paper reframes AI healthcare liability across 31 US cases

A new paper in Nature Health studies 31 US legal cases involving AI tools in healthcare, and argues the complaints point past the treating physician to the insurers, state health departments and care facilities that deploy those systems.

The authors, Gennie Mansi and Mark Riedl of the Georgia Institute of Technology, drew their case pool from three public databases reviewed as of February 2025: the Database of AI Litigation, the Health Litigation Tracker, and the AI, Algorithmic and Automation Incidents and Controversies repository. The systems named across those cases include NaviHealth's nH Predict, Cigna PxDx, EviCore, the Oregon DHS Safety at Screening Tool, NarxCare, the Epic sepsis model, Google Derm Assist, Babylon Health's triage system, IBM's Oncology Expert Advisor and the Veterans Affairs suicide prevention algorithm.

"Patients' medical care relies on a complex web of stakeholders," Mansi and Riedl write, naming physicians, state health departments, health insurers and care facilities among the actors whose decisions shape algorithmic care.

Their frame, in their own words, is a shift "from physician-centered to patient-centered accountability approaches." They propose two moves: changing liability structures to reflect the role of many stakeholders, and designing explainable-AI tools that can serve "advocates, such as legal representatives, who provide critical legal expertise" alongside the clinicians most XAI work currently targets.

The paper is a design-implications argument, not an empirical outcomes study. It does not report how any of the 31 cases were decided, or how often each named system shows up in the pool.