生成式人工智能对数据安全保护的挑战及刑法应对  

The Challenges and the Criminal Law Response of Generative Artificial Intelligence to Data Security

作　　者：聂立泽[1] 王祯 Nie Lize;Wang Zhen

机构地区：[1]中山大学法学院,广东广州510006

出　　处：《河南社会科学》2025年第1期56-63,共8页Henan Social Sciences

摘　　要：随着生成式人工智能在多个领域的应用,数据安全保护面临新的挑战。在数据安全保护的路径选择上,传统模式选择以数据主体为锚定权利的基准,在赋予数据主体相关权利的基础上,以告知—同意原则作为数据主体的权利边界,建立涉及刑法、民法等多个领域的数据安全制度。但不得不承认,伴随人工智能的发展,这种基于数据主体权利的数据保护模式逐渐表现出认知与结构性困难,需要从路径选择上进行转变。相对于以数据主体权利为核心的保护模式,刑法可以从风险的角度考虑,以贝克的风险社会理论作为理论参考,从承认数据主体具有完全理性转变为承认其具有有限理性,将数据安全保护的主要关注点从数据主体的权利保障转移到数据安全风险的预防和管控。此外,在数据安全保护的实质层面上,生成式人工智能侵害数据法益的行为可以概括为行为人利用或针对生成式人工智能实施的行为,以及生成式人工智能在演化、运行过程中“自主”实施的行为。刑法应以风险作为主要考量因素,结合已有规定对上述两种行为进行评价。With the application of generative artificial intelligence in various fields,data security protection faces new challenges.In the path selection for data security protection,the traditional model anchors rights on the data subject,establishing a data security system involving criminal law,civil law,and other areas based on informing-obtaining consent as the boundary of the data subject’s rights.However,it must be acknowledged that with the development of artificial intelligence,this data protection model based on the rights of the data subject gradually exhibits cognitive and structural difficulties,necessitating a shift in approach.Compared to a protection model centered on the rights of the data subject,criminal law can consider from a risk perspective,using Beck’s Risk Society Theory as a theoretical reference.This involves shifting from assuming the data subject possesses full rationality to acknowledging limited rationality,focusing data security protection more on the prevention and control of data security risks rather than solely on safeguarding the rights of the data subject.Moreover,in terms of the substantive aspects of data security protection,actions by generative artificial intelligence that infringe upon data legal interests can be summarized as behaviors carried out by individuals using or targeting generative AI,and actions“autonomously”taken by generative AI during its evolution and operation.Criminal law should primarily consider risk factors and evaluate the two types of behaviors mentioned above in conjunction with existing regulations.

关 键 词：生成式人工智能 风险预防 数据安全 刑法规制 

分 类 号：D9[政治法律—法学]