Artificial intelligence medical scribes promise to reduce physicians' documentation workloads, but research showing that the systems can omit or invent clinical information raises questions about patient safety and malpractice liability.
A 2026 study published in JMIR Medical Informatics evaluated an ambient AI scribe used by 31 physicians to generate 7,545 clinical notes. Physicians assessed a sample of 356 notes, identifying omissions in 18%, invented information in 11.5%, and potentially serious errors in 5.3%.
The researchers classified the serious errors as posing potential harm if left uncorrected, not as documented patient injuries. The study evaluated one AI scribe platform, and its findings do not establish error rates for other products.
AMA Calls for Greater Vendor Accountability
In November, the American Medical Association adopted a policy urging federal and state lawmakers to establish protections for physicians when defective AI products harm patients despite appropriate use and clinical judgment.
The AMA's proposed framework would place responsibility for product-related harm on vendors, manufacturers, or distributors rather than automatically exposing physicians to malpractice liability.
The association also advocates stronger liability insurance requirements and protections against vendor contracts that shift responsibility for defective technology onto physicians.
The AMA's position does not change existing malpractice law. Physicians remain responsible for exercising appropriate clinical judgment, while actual liability depends on applicable law and the circumstances of each case.
California Requirements and Physician Responsibilities
California's AB 3030 requires physician offices and other specified healthcare organizations to notify patients when generative AI produces communications containing patient clinical information, unless a licensed or certified healthcare professional reads and reviews the communication.
The law addresses patient-facing communications and does not replace California's separate consent requirements for recording confidential conversations.
For San Diego physicians adopting AI medical scribes, these distinctions make patient notification, recording consent, and documentation review important elements of practice management.
Reducing Documentation and Liability Risks
Independent physicians and medical groups can establish safeguards before incorporating AI-generated notes into patient records.
Practices should require clinicians to verify diagnoses, symptoms, medications, allergies, and treatment plans before signing AI-generated documentation. They should also examine vendor contracts, patient-consent procedures, recording-retention policies, and malpractice insurance coverage.
Medical groups should establish procedures to identify recurring errors and report problems to vendors.
Research has documented potentially serious AI-scribe errors, but it has not established that these systems have increased malpractice claims.
For San Diego physicians, the challenge is capturing AI's administrative benefits while maintaining accurate medical records, protecting patient privacy, and exercising independent clinical judgment.
