Tokyo court recognizes a human voice as a publicity-rights subject in AI voice-clone suit against TikTok; takedown claim dismissed as moot

What happened

On 30 September 2026, the Tokyo District Court (presiding judge Aya Takahashi) ruled in the suit voice actor Kenjiro Tsuda brought against the Singapore operating company of TikTok (filed November 2025), reportedly Japan's first lawsuit seeking to defend an individual's vocal identity against AI-generated copies. Tsuda alleged an anonymous account posted at least 188 videos from July 2024 to September 2025 with narration generated to imitate his distinctive "lustrous" baritone, averaging ~1.47 million views per video, with the poster earning an alleged ¥500,000–¥750,000 per month; the complaint pleaded publicity-rights infringement and unfair-competition violations.

The court held that a human voice is "as symbolic of one's personality as portraits are": where a voice has customer-attraction power and is used exclusively to exploit that attraction, its unauthorized use can infringe publicity rights — reported as the first Japanese judicial ruling that a voice is a publicity-rights subject. The court, however, dismissed Tsuda's takedown claim: the videos and the posting account had already been deleted during the proceedings, so TikTok had no continuing deletion obligation. The court expressly did not decide whether the AI-generated narration was actually similar to Tsuda's voice. TikTok had argued the narration was merely a "generic male voice." Tsuda's lawyer said the purpose of the suit was fully achieved and the voice-publicity holding is the first of its kind; the Japan Actors' Union (~2,600 members) publicly backed the suit.

Why it may matter to you

Japanese courts had previously left open which rights could protect against AI "digital replicas"; this ruling gives performers a publicity-rights foothold against unauthorized AI voice cloning.

What is not yet known

Whether the AI audio was actually found similar to Tsuda's voice (undecided); how the publicity-rights test applies to training data rather than outputs. These are reports of a judgment, not the judgment text itself.