Tort Liability Analysis of Generative Artificial Intelligence in Judicial Application

An Example of Generative Large Models

Abstract

The judicial application of generative artificial intelligence technology has triggered core legal issues such as subject matter eligibility disputes, definition of service attributes and allocation of tort liability. This paper takes Tencent Dreamwriter case and other typical cases as an entry point to systematically explore the difficulties in determining the tort liability of generative artificial intelligence under the current legal framework, focusing on the disputes over the liability of service providers, the dilemma of determining the negligence caused by algorithmic defects, and the ambiguity of defining the infringement behaviours. Through comparative analysis of domestic and international legislative practices, the study reveals the "instrumental" nature of generative AI and the core of the dispute over its legal nature: service providers, as the controllers of the technology, are required to undertake the legal obligations of data compliance review, generation of content labelling, and filtering of illegal information. The study further proposes a functionalist approach to the determination of tort liability, combining the "equivalent causation theory" with the principle of judicial classification to provide a dynamic framework for the division of liability. At the level of risk regulation, China has formed a governance system that combines policy guidance and technical regulation, and has strengthened algorithmic transparency and content traceability through the Measures for the Labelling of Artificial Intelligence-Generated Synthesised Content. The purpose of this paper is to provide both theoretical support and practical guidance for the judicial discretion and institutional improvement of generative AI infringement disputes.