生成式AI虚假身份信息的风险与应对  

Risks and Responses to Generative AI False Identity Information

作　　者：刘谢慈 虞晨 LIU Xieci;YU Chen(Law School,Hunan University of Technology,Zhuzhou,Hunan 412007,China)

机构地区：[1]湖南工业大学法学院,湖南株洲412007

出　　处：《宜宾学院学报》2025年第3期1-9,共9页Journal of Yibin University

基　　金：国家社会科学基金青年项目“数据要素市场化配置下体育赛事数据财产权实现机制研究”(24CTY013)。

摘　　要：生成式AI正通过不断优化内部对抗学习机制生成具有逼真性、多样性、时效性、规模性等逻辑特性的虚假身份信息。其典型样态包括照片和视频图像、文本合成、语音克隆,由此引发的虚假身份信息风险具有多样性,包括“积极生成型”和“人为操纵型”两种情形。基于其复杂的技术因素,生成式AI虚假身份信息的生成与传播会侵害社会和个体的基本权益,从而在法律定位、责任标准上对相关法律规制发起挑战。鉴于此,一方面通过强化技术赋能过程中风险防控的手段,建立多元主体协作共治机制以应对法律和技术的深度融合;另一方面,创设以生成式AI为责任主体的新型诽谤救济模式,在集体主义视角下构建以“人机交互联合体”为导向的归责方法,构建适用于过渡时期的暂代型诽谤救济模式。Currently,generative AI is generating false identity information with logical characteristics such as realism,diversity,timeliness and scale through continuously optimising internal adversarial learning mechanisms.Its typical patterns mainly include photo and video images,text synthesis,and voice cloning,and the risk of false identity information arising from this is diverse,mainly including“actively generated”and“artificially manipulated”scenarios.Based on its complex technical factors,the generation and dissemination of generative AI false identity information will infringe on the basic rights and interests of the society and individuals,thus challenging the current legal regulation of false identity information in terms of legal positioning and liability standards.In view of this,on the one hand,by strengthening the means of risk prevention and control in the process of technological empowerment,the mechanism of multi-dimensional collaboration and cocontrol of subjects to cope with the deep fusion of law and technology is established;on the other hand,a new type of defamation relief model with the generative AI as the subject of responsibility is created,from the perspective of collectivism,the imputation method guided by“human-computer interaction consortium”is constructed,and the temporary defamation relief model suitable for the transitional period is constructed too.

关 键 词：生成式AI 虚假身份信息 诽谤 

分 类 号：D922.16[政治法律—宪法学与行政法学]