Private Law Theory

ABSTRACT
Tort doctrine allocates liability for machine-caused harm by asking what a reasonable manufacturer could have foreseen at the moment a product left its hands. That question assumes a product whose risk profile is fixed at the point of sale. Autonomous artificial intelligence systems violate this assumption: their decision-making capacities change after deployment through continued training, fine-tuning, contextual adaptation, and interaction with data the original designer never observed. This article argues that negligence and product-liability doctrine have responded to this mismatch inconsistently and, in most jurisdictions, implicitly rather than explicitly, producing causation outcomes that track litigation strategy more than the underlying epistemic reality of how these systems fail. Using a comparative doctrinal method across the European Union, the United States, the United Kingdom, Germany, Japan, and Brazil, the article traces how the temporal reference point of foreseeability is being relocated-largely without express acknowledgment-from the moment of design to the moment of last meaningful system modification. It examines Benavides v Tesla, Inc, the EU’s Product Liability Directive (EU) 2024/2853, the UK’s Automated Vehicles Act 2024, and Brazil’s pending PL 2338/2023 as evidence of this unarticulated shift. The article’s original contribution is a proposed doctrinal reconstruction-the Dynamic Foreseeability and Monitoring Standard-under which a continuing duty to monitor deployed systems generates an independent, judicially cognizable causal pathway distinct from the original design defect. The article evaluates this proposal against institutional capacity constraints, identifies its strongest doctrinal objections, and specifies the evidentiary conditions under which the standard would not apply. AI Disclosure Statement During the preparation of this manuscript, the author used AI-based language and research-assistance tools for drafting support, research organization, literature discovery, analytical structuring, and language refinement, including verification searches for the comparative regulatory and case-law material discussed herein. The author critically reviewed, edited, verified, and assumed responsibility for all legal analysis, argumentation, sources, interpretations, and conclusions contained in the final manuscript.

Salomão De Oliveira, Northon, Artificial Intelligence Liability and Tort Law: Causation, Foreseeability, and Legal Responsibility for Autonomous AI Harm (August 22, 2026).