肖像可识别性法益区分标准下AI融脸侵权的类型化认定

Typological Determination of AI Face-Swapping Infringement Based on the Differentiated Identifiability Standards of Portrait Rights

摘要: AI融脸技术的快速迭代使肖像侵权形态日趋复杂，传统的综合判断法难以应对“去面部特征”场景下的可识别性认定难题。AI融脸侵权的认定不应采取“一刀切”的单一标准，而应区分肖像权所保护的不同法益，按照尊严利益、同一性利益与财产性利益分别适用差异化的可识别性判断标准。尊严利益保护应采用“本人识别”标准，同一性利益保护应采用“熟人确信”标准，财产性利益保护应采用“公众一眼识别”标准。通过对AI换脸模板案、AI恶搞丑化案、AI短剧“神似”明星案等典型案例的类型化分析，揭示了多元标准体系在司法实践中的适用逻辑，据此就AI融脸侵权的制度完善提出建议。

Abstract: The rapid iteration of AI face-swapping technology has rendered the forms of portrait infringement increasingly complex, and the conventional “holistic assessment” approach proves inadequate in addressing the challenge of determining identifiability in scenarios involving “removed facial features”. The determination of AI face-swapping infringement should not adopt a one-size-fits-all uniform standard; instead, it should distinguish among the different legal interests protected by the right of portrait—namely, dignity interests, identity interests, and commercial interests—and apply differentiated identifiability standards accordingly. For the protection of dignity interests, the “rights-holder self-recognition” standard should be applied; for identity interests, the “people familiar with the rights holder” standard should be applied; and for commercial interests, the “general public immediate recognition” standard should be applied. Through a typological analysis of typical cases—including AI face-swapping template cases, AI-based defamatory caricature cases, and AI short drama “striking resemblance” to celebrities cases—this study reveals the applicative logic of the multiple-standard framework in judicial practice, and accordingly puts forward recommendations for the institutional improvement of AI face-swapping infringement regulation.