司法 · AI 声音权利
东京法院保护声音权——但未裁定津田所称的 AI 克隆是否侵权
由于视频已经删除,津田要求删除内容的诉求被驳回。具有重要意义的裁决是认可声音可受肖像权保护,而非认定涉嫌 AI 模仿已经构成侵权。
原语言视频
文字实录 · 正在加载
译自英文原文。
新闻追踪第 4 天
1 篇报道25 家媒体28 个来源
津田健次郎要求删除 TikTok 上涉嫌模仿其声音的 AI 视频,但这一请求被驳回。不过,此案为表演者带来了一项重要认可:他们的声音和面孔一样,也可能受到肖像权的保护。表面上的矛盾有一个直接解释:相关视频和账号当时已经消失,东京地方法院并未裁定这些具体视频是否侵犯了他的权利。4 6
2026年9月30日的裁决,重要之处在于它确认了什么,而非施加了何种处罚。它将表演者身份中具有商业价值的一部分纳入法律保护范围。这既不是法院在本案中下令删除内容,也不意味着法院认定所有未经授权的合成模仿都违法。17
主审法官高桥绫将人的声音视为个人身份的表达,认为它可以比照肖像加以保护。《东京新闻》刊载了她的日文推理过程;以下内容是对该报道所引段落的翻译,并非经完整判决书独立核对的原文。4
一个人的声音和其肖像一样,都是其个性的象征。
《卫报》和日本法律新闻报道将该裁决称为日本司法首次明确承认声音受肖像权保护。这是一项重大的法律进展,但承认某项利益受到保护,与证明该利益遭到侵害,是两个不同的问题。本案回答了前者,后者则未作裁定。1 6
保护有明确的商业边界
适用条件十分具体:未经授权使用表演者的声音,若唯一目的在于利用其声音的商业吸引力——即吸引顾客和促进销售的能力——便可能侵犯肖像权。“唯一”一词并非无关紧要。声音容易辨认且未获许可,本身并不足以证明符合这一标准。17
仍有重要边界尚未厘清。现有报道没有说明法院会如何看待戏仿、评论或其他具体用途。这也不是对合成声音或人工智能的一概禁止。在没有完整判决书的情况下,若把报道所述的法律原则扩展为适用于所有类型模仿的全面规则,还为时过早。
据9月30日的 ORICON 报道,津田的案件涉及一个匿名 TikTok 账号于2024年7月至2025年9月期间发布的188个视频。他指称,这些视频使用了一个 AI 生成的声音副本,而该声音只能是他的。这是他对声音身份的主张,并非对音频制作方式的技术认定。没有经核实的报道指出上传者身份、AI 工具、训练素材或模仿方法。1 14
津田的声音不仅容易辨认,也具有职业价值。他在《咒术回战》中为七海建人配音,在《黄金神威》中为杉元佐一配音;他的作品还包括《游戏王!》。法新社形容他的嗓音“浑厚光润”。对于声音本身就是观众寻求其作品的原因之一的表演者而言,模仿所利用的可能不只是熟悉感,而是更为重要的东西。16 20 24
据《卫报》报道,津田的法律团队将该账号发布的内容形容为“下流”。这是律师的说法,并非司法认定的刑事猥亵内容。同样,不能仅仅因为法院承认声音可以受到保护,就把据称发生的模仿视为已获技术证实。1
删除内容已实现实际目标
诉讼提起后,账号运营者删除了视频,连账号也一并删除。争议内容消失后,法院驳回了要求 TikTok 运营公司将其删除的诉求。因此,驳回并不意味着法院否定了对声音的保护。6
据 Bengoshi.com 报道,津田的律师名为平野敬。他的话译为:“所有视频都和账号一起被删除了,因此我们最初的目标已经完全实现。”现有报道无法确认其姓名的英文拼写,也未说明是否计划上诉。报道也没有证实存在损害赔偿请求或判决,或另行申请的预防性禁令。6
《卫报》9月30日的标题称,被复制的声音“侵犯了他的权利”,这使两者之间的区别变得模糊。ITmedia 当天发布的报道明确指出,法院没有裁定这些视频是否构成侵权。日文详细报道所支持的是更为有限的结论:法院认可声音可以受到保护,但并未对已删除视频的责任作出裁决。1 17
报道所述的账号规模说明了商业问题为何重要。《卫报》称,该匿名账号一度拥有超过200,000名订阅者。这是过去的受众数据,并非仍在运营的账号当前订阅总数;额外核查也未能独立证实这一数字。1
收益数字也需要更正。《卫报》最初的报道将月收入写为“¥500,000 to ¥75,000”。9月30日的 ORICON 报道则写作月50万~75万円,即¥500,000–¥750,000。《读卖新闻》2026年5月23日的报道给出的区间相同。原报道中的上限小了10倍;即便只是把英文中由高到低的区间顺序倒过来,也会得到另一个仍然错误的估算。1 14 15
这些收入仍属估算,并未经过审计。《读卖新闻》称,估算依据是上传者在另一个网站发布的信息。该报道和 ORICON 的描述都不能证明实际付款已获核实,也不构成法院对月收入的认定。将估算乘以视频发布时长,也不能证明总收入已经确定。14 15
《读卖新闻》较早的报道说,从2024年7月至提起诉讼期间,至少发布了188个视频;9月30日的 ORICON 报道则将时间段具体写为2024年7月至2025年9月。后者提供了更具体的发布时间范围。但两种说法都不能确定诉讼的确切提起日期。14 15
日本演员工会支持津田采取法律行动,并感谢他的勇气,同时承认诉讼过程会带来情绪压力。该组织声明称,将审阅正式判决书,之后在其网站上发表评论。声明没有公布发布日期,现有材料也无法证实上诉计划或案件编号。23 18
具体结果意义重大,但范围有限:津田要求删除内容的目标已经实现,法院也认可声音可受肖像权保护。至于另一种有争议的使用是否符合“以商业利用为唯一目的”的标准,仅凭这一认可无法回答。
本报道参考的来源
播出原貌 11 行
- A Tokyo court has recognized a performer’s voice as protected by publicity rights.
- That gives performers a legal basis to protect a sound that attracts audiences and generates income. But actor Kenjiro Tsuda’s case produced neither a finding that the disputed videos infringed his rights nor a court-ordered takedown.
- What does that protection cover when the disputed videos disappeared before the court ruled?
- A voice can carry personal identity just as a likeness does. Presiding Judge Aya Takahashi recognized that connection in the Tokyo District Court’s September 30, 2026 decision. The protection extends beyond what a performer looks like to the sound that can draw customers and promote sales. But the commercial condition is crucial: unauthorized use can infringe publicity rights when it is exclusively directed toward exploiting the voice’s customer-attracting power. Recognition of a protected voice is not the same thing as proving an unlawful use.
- That exclusive-purpose requirement leaves real boundaries unresolved. The decision is not a blanket prohibition on every unauthorized imitation, every synthetic voice, or every use of artificial intelligence. It does not settle how the rule applies to parody, commentary, or other particular uses. The complete judgment is unavailable for independent examination, limiting what can be established about the full reasoning.
- Tsuda’s voice is part of the work audiences know him for. He plays Kento Nanami in Jujutsu Kaisen and Hyakunosuke Ogata in Golden Kamuy, and his credits include Yu-Gi-Oh! His distinctive baritone helps explain the commercial stake: an imitation can attach a recognizable performer’s sound to content he never authorized. The material at issue has not been established by the court as criminally obscene.
- Tsuda’s complaint concerned 188 videos on an anonymous TikTok account. The alleged imitation was an AI-generated voice that could only be his, but that identification remains an allegation, not a technical finding. No verified account identifies the uploader, the AI tool, the training material, or the method used to produce the audio. The posting window was July 2024 through September 2025. An earlier description counted at least 188 videos from July 2024 until filing. Neither establishes the exact lawsuit filing date; the upload period must not be mistaken for it.
- The court left infringement by these particular videos undecided. A claim that the copied voice was judicially found to violate Tsuda’s rights goes beyond the established outcome. The account operator deleted the videos after the lawsuit was filed, and the account itself was removed too. That left nothing for the requested deletion order to remove. The court dismissed the demand against TikTok’s operating company because the material was already gone—not because it rejected protection for voices. The removal achieved the lawsuit’s original practical objective. No damages demand or award, or separate prospective injunction, has been established.
- The account’s historical audience estimate exceeds 200,000 subscribers. That figure has not been independently corroborated, and it is not the audience of an account still operating after deletion. Audience size helps explain the alleged commercial stakes, but it does not prove liability.
- The upper monthly earnings estimate is 750,000 yen, not 75,000 yen. The correct range is 500,000 to 750,000 yen a month. In the Japanese notation, fifty units of ten thousand means 500,000, and seventy-five means 750,000. The mistaken upper figure was ten times too small; reversing the descending range would not fix it. These earnings remain an allegation. The estimate was based on information the uploader published on another website, not independently verified payments, audited financial records, or a judicial income finding. Multiplying it across the posting period would not establish proven total revenue.
- The Japan Actors Union supports Tsuda’s action and thanks him for his courage. Its organizational statement also acknowledges the emotional strain of the proceedings. The decision has been described as Japan’s first express judicial recognition of voice publicity-rights protection. That recognition supplies a legal basis, but whether another disputed use satisfies the exclusive-commercial-exploitation test still requires a separate determination. This case did not make that determination for the removed videos. No appeal plan or case number has been established. The union intends to review the full formal judgment and comment afterward, but no publication date has been announced.
来源
28 个来源25 家媒体2 篇通稿
来源地图
追溯本篇报道的依据。图谱展示其关联;下方链接可打开本文引用的来源。
来源图谱将在您阅读时加载。
译文更新于
修订记录
- r1首次发布。
