Docs Can't Blame AI, Expert

Doctors who use artificial intelligence in patient care still own the decision if something goes wrong, even when the software makes the mistake. (doctors.net.uk) That warning surfaced in January 2026 as medical defence groups and legal commentators argued that a faulty tool is not a stand-alone negligence defense for a clinician. One recent explainer in emergency medicine put it more bluntly: the claim is still most likely to rest on the clinician involved. (doctors.net.uk) (acepnow.com) The legal split is between malpractice and product liability. A patient can sue a doctor, hospital, software vendor, or all three, but malpractice law still turns on whether the clinician met the standard of care for that patient. (acepnow.com) (frontiersin.org) Federal regulators draw the same line in plainer terms: many clinical decision support tools are meant to assist health professionals, not replace them. The Food and Drug Administration’s January 2026 guidance says it is clarifying which clinician-facing decision support functions fall outside device regulation and which remain regulated. (fda.gov) The American Medical Association uses the phrase “augmented intelligence” to stress that these systems are supposed to enhance human judgment rather than substitute for it. The group’s policy work now includes physician liability, transparency, privacy, and governance for AI-enabled tools. (ama-assn.org) That debate is no longer theoretical because use is spreading fast. The American Medical Association said in March 2026 that more than 80% of physicians reported using AI in their professional work, double the rate reported in 2023. (ama-assn.org) Hospitals are building rules around that growth. The American Medical Association’s August 2025 governance toolkit told health systems to set executive accountability, vet vendors, define permitted and prohibited uses, and train staff before rolling out new tools. (fiercehealthcare.com) (acdis.org) Courts have also shown why vendors may not absorb the whole blame. In one case discussed by ACEP Now, judges found the duty of care ran between the patient and the clinicians using an algorithmic heart-risk report, not between the patient and the software developer. (acepnow.com) That leaves clinicians in a familiar position with a newer tool: they can use software to inform care, but they still have to recognize when its recommendation does not fit the patient in front of them. (ama-assn.org) (doctors.net.uk)