生成式人工智能规制中的义务主体与义务分配——以ChatGPT为例  

Obligatoion Subjects and Obligation Allocation in Generative AI Regulation——A Case Study of ChatGPT

作　　者：万斌 WAN Bin(School of Law,Anhui University,Hefei Anhui 230039,China)

机构地区：[1]安徽大学法学院,安徽合肥230039

出　　处：《太原理工大学学报（社会科学版）》2024年第3期29-35,44,共8页Journal of Taiyuan University of Technology(Social Science Edition)

基　　金：安徽省高校人文社会科学重大研究项目“安徽省‘一带一路’参与企业合规风险防范法律问题研究”(SK2021ZD03)。

摘　　要：以ChatGPT为代表的生成式人工智能在给人类带来机遇的同时也产生了数据来源的侵权风险与生成内容的应用风险。目前我国对生成式人工智能的规制体系存在义务主体不清晰和义务分配不合理的问题。在义务主体上,借鉴欧盟《人工智能法案》的规定,应当区分技术支持者、服务提供者与使用者三类义务主体。在义务分配上,为了避免过度防范风险而阻碍技术发展,应当要求服务提供者负担《生成式人工智能服务管理暂行办法》规定的义务并附加风险管理制度的义务;技术支持者仅对预训练数据负责并采取技术措施避免悖伦内容生成;使用者则需合理使用生成内容,专业使用者应负担生成内容的披露义务。Generative artificial intelligence represented by ChatGPT not only brings opportunities to humans, but also generates the infringement risk of data source and the application risk of generated content. At present, the regulation system for generative artificial intelligence in China has the problems of unclear obligation subjects and unreasonable obligation allocation. In terms of obligatory subjects, drawing on the provisions of the EU's Artificial Intelligence Act, three types of obligation subjects should be distinguished: technical supporters, service providers, and users. In terms of obligation allocation, in order to avoid excessive risk prevention and hinder technological development, service providers should be required to bear the obligations stipulated in the Interim Measures for the Administration of Generative Artificial Intelligence Services and additional obligations of the risk management system;The technical supporter is only responsible for the pre-training data and takes technical measures to avoid the generation of unethical content;Users should make reasonable use of the generated content, and professional users should bear the obligation to disclose the generated content.

关 键 词：生成式人工智能 ChatGPT 法律规制 服务提供者 使用者 

分 类 号：DF0[政治法律—法学理论]