Nature Health analyzes 31 US lawsuits over medical AI harm

Two Georgia Institute of Technology researchers read 31 US lawsuits over healthcare AI and came back with the same complaint the plaintiffs kept making: the doctor is not always the person who caused the harm.

In Nature Health, Gennie Mansi and Mark Riedl argue that patient care "relies on a complex web of stakeholders," listing physicians, state health departments, health insurers and care facilities, and that centring the physician misses where the AI tool actually goes wrong. Their case corpus draws on the Database of AI Litigation at George Washington University Law, Georgetown's O'Neill Institute litigation tracker, and the AIAAIC incidents database.

Three systems recur: Oregon DHS's Safety-at-Screening tool, the Babylon Health triage system, and nH Predict. The authors write that "patients have had no option but to seek legal recourse for harms."

From that reading the paper proposes two moves. One is to restructure liability so it reflects the role of many stakeholders in shaping how AI tools impact patient care. The other is "designing tools that can help advocates, such as patients' lawyers, provide critical legal expertise," treating legal recourse itself as a design problem.

The paper is an analysis, not a ruling. It reads the cases and their patterns; it does not report which of the 31 have resolved, or how the FDA's Software as a Medical Device pathway would interact with a stakeholder-wide liability regime.