论“深度伪造”智能技术的一体化规制

Comprehensive Regulation on the “Deepfakes” Technology

"深度伪造"是利用深度学习技术中的"对抗生成网络"模型形成的智能视频处理技术。"换脸视频"是该技术最为典型的应用。与原有技术相比,该技术具有高度真实性、泛在普适性和快速演化性等特征。对其滥用不仅将侵犯公民的个人权利,破坏社会稳定与国家安全,甚至还可能消解社会共同的信任。现有技术规制中采用的"溯源防伪"与"反向破解"的方案无法有效应对。现有法律规制中民事、行政、刑事责任因其"事后性"也难以全面覆盖。下一阶段,要在现有技术规制与法律规制基础之上构建"平台-制作者-受众"三位一体规制体系,强化平台的审查责任、明确视频制作者的说明义务并培养公民在"后真相时代"的信息素养。

"Deepfakes"is a new intelligent video processing technology by using the"generative adversarial network"model of deep learning technology.The"Deepfakes"videos formed by this technology have the characteristics of high authenticity,ubiquitous popularity and rapid evolution.Its abuse will not only infringes upon the individual rights of citizens,undermines social stability and national security,and may even dissolve the common trust of the society.The schemes of"tracing"and"cracking"adopted in the existing technical regulations cannot be effectively dealt with the risks causing by"Deepfakes".The civil,administrative and criminal liabilities in the existing legal regulations are difficult to be fully covered the foresaid risks.In the next stage,on the basis of existing technical regulations and legal provisions,the"platform-producer-audience"regulatory system should be established to strengthen the censorship responsibility of the platform,clarify the clarification obligation of video producers and cultivate citizens’information literacy in the"post-truth era".