Tokyo Court Rules a Voice Is Legally Protected for the First Time in an AI Cloning Case

October 2 (IT Home) — On 30 September local time, according to AFP, a Tokyo court upheld the claim of the well-known Japanese voice actor Kenjiro Tsuda in his case against TikTok, ruling that an AI imitating his deep baritone voice without permission amounted to an infringement of rights. It was the first time a Japanese court has recognised that a person’s voice is protected by law.

Tsuda has worked as a voice actor for more than 30 years, providing voices for a great many films, animated series and video games; among his best-known roles is Seto Kaiba in Yu-Gi-Oh!. He said that an anonymous TikTok account had used an AI clone of his voice as narration in several videos uploaded between July 2024 and the second half of last year. The court partly upheld his claim, but because the account in question had already been deleted, it did not support his request for TikTok to take the videos down.

In its ruling, the court stated that using an actor’s voice without their permission should be regarded as an infringement of image rights. This is also the first case of its kind to be heard in Japan.

Tsuda argued that the AI narration in the videos had his distinctive low, “rich” voice, which was enough to conclude that the voice was imitating him. His legal team said the account in question once had more than 200,000 followers, and that it was able to attract so much attention precisely because having a voice like that of a famous voice actor narrate “suspicious, vulgar content” gave it a novel appeal.

Court records show that TikTok argued the videos simply used a “generic male voice,” and that whether it really resembled Tsuda’s was disputed.

The Japanese government is actively promoting AI development as a matter of national strategy, but anime voice actors are growing increasingly worried that creators still lack adequate protection, because their work and their personal likeness are often used to train AI. Last year, the Japanese government asked OpenAI not to use anime and manga content — which it called “irreplaceable treasures” — to train the Sora 2 generative model in a way that infringes copyright.