Private Law Theory

ABSTRACT
Clinical artificial intelligence has moved from pilot deployment to routine use in radiology, pathology, triage, and diagnostic support, yet malpractice doctrine still asks a single question designed for a single actor: did the treating physician breach the standard of care of a reasonably prudent practitioner? This article argues that the persistence of that question, unmodified, produces a structural liability gap whenever an AI system contributes to a diagnostic or treatment error. Drawing on doctrinal, comparative, and institutional analysis across the United States, the European Union, Germany, the United Kingdom, Japan, and Brazil, the article shows that four candidate defendants – the treating physician, the healthcare institution, the device manufacturer, and the AI developer – each escape meaningful accountability under existing frameworks for a different reason: physicians are protected by deference doctrines calibrated to human colleagues, institutions diffuse responsibility through credentialing and procurement layers, manufacturers invoke learned-intermediary and software exclusions from product liability, and developers rely on continuous post-market model drift to sever the causal chain between design and harm. The article’s original contribution is the Clinical Algorithmic Liability Attribution Standard (CALAS), an institutional framework that reallocates the burden of causal proof according to a system’s autonomy tier and mandates a shared, apportioned liability pool financed jointly by developers and institutional deployers rather than litigated de novo in each malpractice claim. The analysis proceeds through comparative doctrinal review, an institutional and empirical account of AI-enabled device authorization and clinical uptake, and jurisprudential analysis of the human-oversight standards that have emerged from adjacent algorithmic-accountability litigation. The article concludes that without a dedicated attribution standard, the liability gap will not close through incremental judicial adaptation of existing malpractice and product liability doctrine, because those doctrines were built on an assumption-a single, identifiable, professionally licensed decision-maker-that AI-assisted medicine no longer satisfies.

Salomão De Oliveira, Northon, Medical Malpractice, Artificial Intelligence and Physician Liability: Who Is Responsible for AI-Assisted Medical Errors? (August 10, 2026).