Last week, the Japan Actors' Association (JAU) released its first guidelines on generative AI use, setting regulatory requirements targeting the unauthorized use of AI to replicate actors' voices. The guidelines apply to actors' voice-imitating audio generated through technologies such as speech synthesis, deep learning, and machine learning, and require the relevant parties to obtain the actor's own consent at every stage of the voice replication process. If AI companies use actors' voices for model training, they must obtain the performers' permission; AI voices generated with authorization are regarded as a kind of "performance," and companies should pay actors remuneration or royalties.
Japanese voice actors have long opposed unauthorized voice theft. In 2024, several voice actors formed an organization to protest the misuse of AI voices, pointing out that existing copyright laws provide insufficient protection. Kenjiro Tsuda, who has voiced characters in Attack on Titan, Chainsaw Man, and Yu-Gi-Oh!, once sued TikTok over the proliferation of videos on social media that used his voice without authorization, but the case was dismissed after the relevant videos were taken down.
At the guidelines' press conference, Masako Nozawa, known for playing Son Goku in Dragon Ball, said that using her voice without permission is no different from theft, and that voices and performances should be respected—even a small amount of use requires prior consent. She also noted that AI-replicated voices lack emotional expression.
The Japan Actors' Association has about 2,600 voice actor and actor members, and these guidelines are among Japan's first industry standards for AI voice recording. Although Japanese society generally shows a high level of acceptance toward AI, voice actors widely worry that the misuse of their voices will infringe on their labor and squeeze employment opportunities. As the Japanese government continues to explore a regulatory framework for generative AI, the guidelines also provide an industry reference for AI voice licensing, training permissions, and revenue distribution.