Tokyo court says a performer's voice can carry publicity rights, but dismisses TikTok deletion claim

Hero illustration: a generic civic entrance, not the Tokyo District Court building or a scene from this case. AI-generated conceptual illustration by BIG CHANGE.

A Tokyo District Court judgment on September 30 gives voice performers a conditional legal argument against commercial exploitation of their voices. It did not decide whether the narration in the 188 TikTok posts at issue sounded like actor and voice performer Kenjiro Tsuda, used his voice, or infringed his rights. The court dismissed Tsuda's requests to remove the posts because their data had already been deleted.

The big change

The court said a person's voice can symbolize identity like a likeness. Unauthorized use of a performer's voice can infringe publicity rights when the voice attracts customers and the use is solely aimed at exploiting that appeal. The judges did not decide whether these clips met that test.

Performers challenging commercial voice imitation can now cite that judicial statement, but they still need evidence tying a disputed voice and its use to the claim. Tsuda's deletion request ended after the posts were removed, leaving similarity, infringement and TikTok's responsibility for the posts unresolved.

Why the deletion request failed

Tsuda sued TikTok Pte. Ltd. for removal under Japan's Unfair Competition Prevention Act and an asserted publicity right. The nine-page judgment says an unidentified account posted 188 videos from July 2024 to September 2025, pairing images and subtitles with narration about urban legends and paranormal topics. Tsuda argued that the narration resembled his voice. The account's linked page said its AI audio used the voice of a friend skilled at imitating him, rather than recordings of Tsuda. The judgment records that statement as part of the evidence; it does not verify how the audio was made.

The poster deleted the account on May 29, 2026. The court found that no restoration request followed and that the account's data, including the disputed posts, was removed from TikTok's servers after June 28. Even if TikTok could otherwise have had a duty to delete the posts, there was nothing further for it to remove. The court dismissed both deletion claims and expressly declined to decide the remaining issues. This was a dismissal of the requested remedy after deletion, not a ruling that the posts were lawful. Tsuda had not sought damages.

What the court said about voice

Before addressing deletion, the judges wrote that a person's voice, like their likeness, can symbolize their identity. Their formulation was specific: unauthorized use of a performer's voice may infringe publicity rights if the voice has customer-attracting power, such as helping sell a product, and the use is solely aimed at exploiting that power. Those are English renderings of the judgment's terms 顧客吸引力 and 専ら. The court did not say that every imitation, every AI-generated voice, or every unauthorized use satisfies those conditions.

Nor did the court decide whether Tsuda's voice was used in these videos, whether the sound was sufficiently similar, whether it attracted viewers, whether the poster's purpose met that test, or whether TikTok itself had deletion liability. Tsuda's side submitted a voice comparison analysis, while TikTok disputed similarity and said the audio came from a third party's voice. Those remain opposing positions in this case, not findings of infringement.

The legal question predates this ruling. Japan's Ministry of Justice published an August 2026 interpretive report on civil liability for unauthorized use of likenesses and voices, including generative AI uses. The ministry describes it as an interpretation of existing law and case principles. The district court's statement now gives that issue a judicial formulation, although the case's deleted-post outcome prevented a decision applying it to Tsuda's allegations.

At a September 3 seminar at Waseda University, Tsuda's lawyer Kei Hirano described the difficulty of proving voice similarity in a process centered on written records, Bengo4 News reported. Hirano said Tsuda's team had submitted analysis using multiple deep learning models and the program code. That is the lawyer's account of the evidence, not a court finding that the voices matched. Separately, Bengo4 reported Hirano's post-judgment statements that Tsuda did not plan to appeal and had sought removal rather than money.

For performers and platforms, the practical boundary is clear but limited: a commercially valuable voice can be relevant to a publicity-rights claim, while liability for a particular imitation still depends on evidence and the legal test. This article explains the published judgment; it is not legal advice.