论生成式人工智能学习端的个人信息分级同意规则  

On the Graded Consent Rule for Personal Information at the Learning Stage of Generative Artificial Intelligence

作　　者：朱晓峰[1] 袁子烨 ZHU Xiaofeng;YUAN Ziye

机构地区：[1]中央财经大学法学院 [2]中国政法大学中欧法学院

出　　处：《西北工业大学学报(社会科学版)》2025年第2期136-143,共8页Journal of Northwestern Polytechnical University(Social Sciences)

基　　金：国家社会科学基金后期资助项目“个人信息侵权责任认定规则研究”(24FFXB048)。

摘　　要：生成式人工智能学习端的数据抓取算法存在超目的处理个人信息等问题,未在个人信息处理之前明确告知信息主体处理目的并获得同意,架空了既有的知情同意规则,削弱了信息主体对个人信息的知情权和自决权。为此,应在坚持信息主体在个人信息处理中的中心地位、保留现有知情同意规则基本框架的前提下,引入分级同意理论,修正《个人信息保护法》中的知情同意规则,构建动态化、层级化的分级同意规则。具体来讲,应以现行法上对同意方式的分级为基础,运用解释论的方法将信息主体的同意细化为严格同意、强同意、弱同意和不需要同意等四级,并将不同类型的个人信息分别置于这四级同意规则中处理。The data collection algorithms at the learning stage of generative artificial intelligence face issues such as the overreach in processing personal information.These algorithms fail to clearly inform data subjects of the processing purposes and obtain consent before handling personal information,thereby bypassing existing informed consent rules and undermining the data subjects'rights to be informed and to make decisions about their personal data.To address this,it is necessary to uphold the central role of data subjects in personal information processing and retain the basic framework of the current informed consent rules.Simultaneously,the theory of graded consent should be introduced to amend the informed consent rules in the Personal Information Protection Law,establishing a dynamic and hierarchical graded consent framework.Specifically,based on the current legal classification of consent methods,an interpretative approach should be used to refine data subjects'consent into four levels:strict consent,strong consent,weak consent,and no consent;different types of personal information should be handled according to these levels.

关 键 词：生成式人工智能 DeepSeek 知情同意 个人信息 分级同意 

分 类 号：D923[政治法律—民商法学] D922.16[政治法律—法学]