人工智能深度伪造技术的法律风险防控

Research on Legal Risk Prevention and Control Countermeasures of Deepfake

作为生成式人工智能的深度伪造技术给社会带来了新的风险与挑战。从类型学上看,深度伪造技术同时引发了基于个体权利谱系的微观法律风险、基于市场经济秩序的中观法律风险和基于国家公共安全的宏观法律风险,并依此形成了多样化和梯次性的法律风险格局。为有效化解风险,各国存在“自下而上型”分散治理模式、“自上而下型”垂直治理模式以及“多方参与型”合作治理模式。我国可通过“软法+硬法+算法”相结合的综合治理模式,实现对前述法律风险的有效防控:在软法层面,完善科技伦理制度,明确科技伦理的实体规范和程序规范;在硬法层面,通过“风险预防”与“风险控制”,强化风险防控的刚性约束;在算法层面,实现从样本数据采集到深度伪造内容发布的全过程技术监督,发挥技术的规范功能。

The Deepfake has realized the technological revolution from continuous innovation to destructive innovation,and has produced both positive and negative effects in its application.From a typological point of view,the Deepfake simultaneously triggers micro legal risk based on individual rights spectrum(type I),medium legal risk based on market economic order(Type II)and macro legal risk based on national public security(type III),and thus forms a diversified and hierarchical legal risk pattern.In order to effectively resolve risks,the United States,the European Union and China have respectively formed the"bottom-up"decentralized governance model,the"top-down"vertical governance model and the"multiparticipation"cooperative governance model,and each presents distinct advantages and disadvantages.In view of this,China can achieve effective prevention and control of the above-mentioned legal risks by forming a comprehensive governance model combining"soft law+hard law+algorithm":at the level of soft law,we should perfect the system of science and technology ethics and clarify the substantive and procedural norms of science and technology ethics;at the level of hard law,the rigid constraints of risk prevention and control could be strengthened through the coordination of the two subsystems of risk prevention and risk control;at the algorithm level,we should realize the whole process technical supervision from sample data collection to deepfake content release,and give full play to the normative function of technology.