生成式人工智能侵權責任認定路徑探析

中圖分類號：D923；D922.17 文獻標志碼：A 文章編號：1003-5168(2025)12-0119-07

DOI:10.19968/j.cnki.hnkj.1003-5168.2025.12.023

Analysison the Path ofDetermining TortLiability of Generative Artificial Intelligence

YANG Meng (Nanjing University of Science and Technology,Nanjing 21oo18, China)

Abstract: [Purposes] Aiming at the infringement problems that may be caused by the use of existing generativeartificial inteligence technology,especiallythepotential threat to civil rights,thisstudyaims to build a reasonable framework for the determination of tort liability,explore the tort liability mechanism applicable to generative artificial intelligence system,and focus on analyzing the causation analysis and fault liabilitystandards,so as to provide scientific basis for relevant legislative practice and ensure a balance between technological development and rights protection.[Methods] Based on the existing cases, this paper analyzes the typical structure of generative artificial intelligence infringement,and examines the imputationprinciple of generative artificial intelligence infringement according to product liability and general fault liability.Through the analysis of legal dogmatics,it is concluded that artificial intelligence infringement is more in line with the connotation of fault liability.[Findings] ① The technical core of generative artificial intelligence is embodied in dynamic interactive services,relying on continuous data updates and algorithm iteration,and lacks the homogenization and risk dispersion characteristics of traditional products. ② The content generated by artificial intelligence is a virtual risk,which is incompatible with the "physical defect" element in the "Product Quality Law". ③ The legislative practice clearly defines generative artificial intelligence as "service"and denies the path of product liability. [Conclusions] The liability attribution mechanism should be constructed with fault liability as the core: When the user is not at fault,the principle of presumption of liability should be applied,and the provider should bearthe burden of prooffor technical compliance;When the user is atfault,the provider shall fulfill strict risk prevention andcontrol obligations and be held jointly andseverally liable with theuser.To address the "algorithmic black box" problem,the "high probability" causation standard is adopted to reduce the burden of proof for the plaintiff,while alowing the provider to be exempted from liability through technical counter-evidence.

Keywords: tort liability; generative artificial intelligence; product liability; fault liability

0 引言

自2018年OpenAI公司推出了風靡一時的ChatGPT語言模型以來，人工智能以席卷之勢在社會各個領域快速布局。（剩余11247字）