人工智能合成语音二创短视频的权利冲突与法律规制

Rights Conflicts and Legal Regulation of AI-Synthesized Voice-Based Derivative Short Videos

摘要: 随着生成式人工智能与短视频二次创作深度融合，AI语音二创在降低创作门槛的同时，也引发声音人格权益、著作权及个人信息权益等复合冲突。本文结合实际案例与司法裁判，围绕训练、生成、传播三个环节分析法律边界，并以人格识别利益、数据利用利益和商业开发利益构建“三阶法律结构”。在此基础上，文章以“可识别性”为声音人格利益判断的重要边界，结合著作权合理使用、训练数据合法利用、平台责任与风险分级提出规制路径，以期兼顾人格保护、技术创新与二创产业发展。

Abstract: With the deep integration of generative artificial intelligence and short-video secondary creation, AI voice-based secondary creation has lowered the threshold for creation while also giving rise to complex conflicts involving voice-related personality rights and interests, copyright, and personal information rights and interests. Drawing on real-world cases and judicial decisions, this paper analyzes legal boundaries across the three stages of training, generation, and dissemination, and constructs a “three-tier legal structure” comprising personality identification interests, data utilization interests, and commercial exploitation interests. On this basis, “identifiability” is taken as an important boundary for determining voice-related personality interests. In conjunction with copyright fair use, lawful use of training data, platform responsibility, and graded risk governance, the paper proposes regulatory pathways aimed at balancing personality protection, technological innovation, and the development of the secondary-creation industry.