论生成式人工智能服务提供者的法律地位及其责任——以ChatGPT为例  

On the Legal Status and Responsibilities of Generative Artificial Intelligence Service Providers——Taking ChatGPT as an Example

作　　者：徐伟 Xu Wei

机构地区：[1]上海政法学院上海司法研究所,上海201701

出　　处：《法律科学（西北政法大学学报）》2023年第4期69-80,共12页Science of Law：Journal of Northwest University of Political Science and Law

基　　金：浙江省哲学社会科学规划课题(20NDQN271YB)“AR技术成果版权法保护研究”。

摘　　要：生成式人工智能行为在技术上的特殊性,对侵权责任带来了一系列影响,包括行为主体的复杂化、加害行为的智能化、损害后果的不确定、因果关系的多元化和过错认定的新颖化。生成式人工智能服务提供者并非搜索链接服务提供者,也不宜被认定为网络内容服务提供者,而应是一种新型网络服务提供者。生成式人工智能服务提供者过错的认定需从生成维度的过错和移除维度的过错来分别加以判断。在生成维度,生成式人工智能服务提供者对自动生成的内容原则上不负有事前审查义务;在移除维度,应基于事实层面生成式人工智能服务提供者是否能以合理的成本和较高的准确度来避免侵权内容的再次生成来决定其是否负有移除义务。鉴于技术仍处于快速发展中,生成式人工智能服务提供者侵权责任的判断标准也需基于技术的发展而适时作出调整。The technical characteristics of generative Al lead to a series of impacts on tort liability,including the complexity of the subject,the intellectualization of the tort behavior,the uncertainty of the damage,the diversification of the causation,and the novelty of the faults.An generative AI service provider is not an information location service provider,nor should it be identified as an internet content service provider,but as a new type of internet service provider.The determination of fault for an generative AI service provider should be divided into two dimensions:fault in the generation and fault in the takedown.In the generation dimension,basically,an generative AI service provider is not obligated to review the generative Al in advance.In the takedown dimension,whether an generative AI service provider is obligated to take down should be determined on the basis of whether it can avoid generating infringing content again with reasonable cost and high accuracy at the factual level.Given the rapid development of technology,specific rules of the tort liability of the generative Al service provider should be timely adjusted in light of the development of the technology.

关 键 词：生成式人工智能服务提供者 ChatGPT 算法 通知移除 

分 类 号：D916[政治法律—法学]