Physician Predicts AI Lawsuits

Doctors are starting to ask a new malpractice question: what happens when the software spots a problem and the clinician does not. The debate accelerated after dermatologist Matthew Zirwas wrote that he expects lawsuits in which “AI got it right, provider didn’t.” (sotwe.com, medicaleconomics.com) The legal issue turns on clinical decision support, a category of software that suggests diagnoses, flags risks, or summarizes visits for health professionals. In January 2026, the Food and Drug Administration updated its guidance on which decision-support tools it regulates and which it treats as excluded “Non-Device” software under the 21st Century Cures Act. (fda.gov) That distinction matters because many tools now entering exam rooms are advisory, not autonomous: they surface information, but the doctor still makes the call. The Food and Drug Administration said its 2026 guidance covers software intended for health care professionals and gives examples separating regulated device software from excluded support tools. (fda.gov) U.S. malpractice law has not yet produced a clear courtroom test for these cases. Medical Economics reported in October 2025 that The Doctors Company had seen no documented U.S. malpractice case in which artificial intelligence was central to the claim, even as practices rolled out ambient scribes, triage systems, and imaging tools. (medicaleconomics.com) The pressure is building because physician use is climbing fast. An American Medical Association survey cited in the group’s August 2025 guidance said nearly 70% of physician respondents used AI tools in 2024, up 38% from the previous year. (acdis.org) As adoption rises, the standard of care can move with it. Medical Economics reported that courts are increasingly invited to weigh evidence-based guidelines and contemporary practice, raising the possibility that ignoring a widely used, clinically validated tool could someday be framed as unreasonable care. (medicaleconomics.com) The American Medical Association has not said physicians should hand decisions to algorithms. Its standing policy calls for clinically validated, transparent systems, physician input in design and deployment, and legal oversight that addresses liability, bias, privacy, and patient safety. (ama-assn.org) The group’s 2025 governance advice gets more operational inside hospitals and clinics. It recommends written policies covering approved uses, accountability and oversight, retention of AI-generated information and visit recordings, transparency to patients and clinicians, and recurring training. (acdis.org) That paperwork matters in both directions. If a clinician follows a flawed recommendation, records can show what the tool produced and who reviewed it; if a clinician rejects a recommendation, the chart can show the medical reason for overriding it. (acdis.org, medicaleconomics.com) The first wave of AI health litigation is already arriving on adjacent issues, even before a landmark “AI was right” malpractice case. In April 2026, a lawsuit in federal court in Northern California challenged health systems’ use of an ambient documentation tool over privacy and recording claims, showing how legal exposure can start with the workflow around AI, not just the diagnosis itself. (hipaajournal.com) Zirwas’s prediction lands in that gap between fast adoption and unsettled law. For now, the safest assumption in medicine is that artificial intelligence may advise, but a human still owns the decision and the record explaining it. (sotwe.com, ama-assn.org)