生成式人工智能侵权归责原则的立场抉择与规范构造——以全国首例AI“幻觉”侵权案的裁判逻辑为中心

Stance Selection and Normative Construction of Liability Principles for Generative AI Infringement—Centered on the Judicial Logic of China’s First Case of AI “Hallucination” Infringement

摘要: 生成式人工智能侵权归责原则的确定，是当前侵权法理论回应技术挑战的核心议题。全国首例因AI“幻觉”引发的侵权纠纷案，为这一争议提供了重要的司法实践样本。该案判决明确将生成式人工智能服务定性为“服务”而非“产品”，适用《中华人民共和国民法典》(后文简称《民法典》)第1165条第一款的一般过错责任原则，并从动态系统论出发构建了三层注意义务体系。然而，该判决所选择的归责方案与学界主张的过错推定说之间存在理论张力。本文以该案裁判逻辑为中心，系统检视生成式人工智能侵权的行为特性与归责困境，比较分析一般过错责任、无过错责任与过错推定责任三种方案的适用效果。研究认为，一般过错责任在举证分配层面存在结构性缺陷，过错责任则面临规范适用上的体系障碍，而过错推定责任在举证平衡、利益协调与制度激励方面具有比较优势。在此基础上，本文提出以过错推定为核心的规范构造方案，包括责任主体的类型化界分、举证责任的合理分配以及推翻推定过错的抗辩事由设计，以期在受害人救济与技术创新之间实现动态平衡。

Abstract: Determining liability principles for infringement caused by generative artificial intelligence is a central issue in current tort law theory’s response to technological challenges. The first nationwide case involving an AI “hallucination” that triggered an infringement dispute provides a significant judicial precedent for this debate. The court ruling clearly classified generative AI services as “services” rather than “products”, applied the general principle of fault liability under Article 1165, Paragraph 1 of the Civil Code of the People’s Republic of China (hereinafter referred to as the Civil Code), and established a three-tier duty-of-care framework based on dynamic systems theory. However, the chosen liability approach in this judgment exhibits theoretical tension with the academic consensus favoring presumed fault liability. Centered on the reasoning of this case, this article systematically examines the behavioral characteristics and liability challenges associated with generative AI infringement, and comparatively analyzes the practical effects of three liability models: general fault liability, strict liability, and presumed fault liability. The study concludes that general fault liability suffers from structural flaws in burden allocation, while fault-based liability faces systemic barriers in legal application; in contrast, presumed fault liability offers comparative advantages in balancing evidentiary burdens, coordinating interests, and incentivizing innovation. Based on these findings, the article proposes a regulatory framework centered on presumed fault liability, including typological distinctions among liable parties, rational allocation of the burden of proof, and design of defenses to rebut presumed fault, aiming to achieve a dynamic balance between victim compensation and technological advancement.