生成式人工智能生成内容标识义务的反思

Reflections on the Labeling Obligations for Generative AI-Generated Content

摘要: 生成式人工智能生成内容具有高仿真性、易传播性与识别困难性，可能引发人格权益侵害、虚假信息扩散及社会秩序风险。标识义务是区分人工智能生成内容与一般内容的重要规制工具。我国现行制度已初步形成以服务提供者为核心的标识义务体系，但仍存在“显著标识”概念模糊、义务范围不清、主体责任分配不均的问题。对此，应细化标识标准，适度扩展服务使用者的义务，并完善行业自律机制。

Abstract: Content generated by generative artificial intelligence is highly realistic, easily disseminated, and difficult to identify. It may give rise to risks such as infringement of personality rights and interests, the spread of false information, and threats to social order. Labeling obligations constitute an important regulatory tool for distinguishing AI-generated content from ordinary content. China’s current legal framework has preliminarily established a labeling obligation system centered on service providers. However, problems remain, including the ambiguity of the concept of “conspicuous labeling”, the unclear scope of obligations, and the uneven allocation of responsibilities among relevant subjects. Accordingly, labeling standards should be further refined, the obligations of service users should be moderately expanded, and industry self-regulation mechanisms should be improved.