深度伪造合成“非真实音视频”的规制现状与完善建议——以《网络音视频信息服务管理规定》为切入点

Regulation Status and Improvement Suggestions of Deepfake“Unreal Audio and Video”——Take the management regulations on network audio and video information service as the starting point

“深度伪造”是一种以人工智能深度学习算法为基础的“非真实音视频”合成技术。依托于“生成对抗网络”,“深度伪造”所合成的“非真实音视频”作品具有高逼真度,若被不法利用可能引发严重危害后果。为抑制“深度伪造”的技术风险,国家网信办牵头制定了《网络音视频信息服务管理规定》,规定了标明合成字样、停止传输、部署鉴别机制等规制措施,并划定该算法的适用禁区。不过,上述规制措施仍存在网络平台负担义务过重、缺乏具体责任承当方式等问题。为克服现有规制措施的局限,应当从明确标识“合成”的具体方式、优化监管模式、细化法律责任承担方式等方面入手,寻求合成“非真实音视频”作品法律规制的进一步完善。

“Deepfake”is an“unreal audio and video”synthesis technology based on artificial intelligence deep learning algorithm.Relying on the“generation confrontation network”and“deepfake”,the“unreal audio and video”works synthesized by“deepfake”have high fidelity.If they are illegally used,they may cause serious harmful consequences.In order to curb the technical risk o“f deepfake”,Cyberspace Administration of China took the lead in formulating the regulations on the management of network audio and video information services,stipulating regulatory measures such as marking synthetic words,stopping transmission and deploying authentication mechanism,and delimiting the restricted area for the application of the algorithm.However,there are still some problems in the above regulatory measures,such as the heavy burden of obligations on the network platform and the lack of specific responsibility bearing methods.In order to overcome the limitations of the existing regulatory measures,we should seek the further improvement of the legal regulation of synthetic“unreal audio and video”works from the aspects of clearly identifying the specific way of“synthesis”,optimizing the regulatory mode and refining the way of legal responsibility.