Japan actors union publishes first AI voice guidelines

Nippairen, the cooperative union that represents Japan's actors and voice performers, published its first formal guidelines on generative-AI use of voices on October 15, 2026, setting voluntary rules that require a performer's prior permission before a voice is trained on, generated or used, payment when an AI-generated voice is used, fresh consent and extra pay when a performance is dubbed into another language by AI, and the permission of a dead performer's heirs before that voice is recreated.

The union, formally the Japan Actors Union (日本俳優連合), announced the guidelines at a press conference in Tokyo on October 15, according to its announcement page and the press release it issued on the PR Times wire. The document, titled 音声に関する生成AI利用ガイドライン (Guidelines on Generative-AI Use of Voices), is marked version 1.0, dated October 15, 2026, and runs to 22 articles, according to the full PDF. The union's release puts its membership at about 2,600 actors and voice actors.

The guidelines are voluntary industry rules, not law. ITmedia AI+ reported that the document has no legal binding force and that the union's first aim is to establish it as an industry standard. The union gives its reason in the guidelines: Japan has no law that specifically protects a person's voice, only case law and debate around likeness and publicity rights, a gap the union calls a legal vacuum. It asks production companies, distribution platforms, AI developers and users to follow the rules while it presses the government to legislate voice rights.

The union singles out four provisions. Part 3, Articles 7 to 9 require prior permission from the owner of a voice at every stage: training and input, generation and construction of a voice model, and publication and use. Part 5, Article 16 defines AI-generated use of a voice as a new form of performance and requires payment, with royalties tied to the scope of use. Part 4, Article 15 requires fresh permission and additional payment when a performance is dubbed into another language by AI. Part 2, Article 6 requires the strict permission of the heirs or the rights holder before a deceased performer's voice is recreated. The article numbers the union lists on its announcement page match the table of contents in the PDF.

Further terms, reported by KAI-YOU, include separate written consent for each of the three stages, notice to the performer at least seven days before a new model is trained and 48 hours before a voice is generated with an existing model, a ban on blanket buyout contracts that grant future use in advance, and a default cap of one year on the term of consent. Pay is split into categories separate from a normal performance fee: a training and development fee, a generated-voice usage fee, an additional fee for advertising use, and a multilingual fee.

At the conference, the union's managing director, Shimada Bin, a voice actor who plays Broly in the Dragon Ball series, said he had heard an AI clone of Broly's voice singing Ado's song Ussēwa, according to the entertainment outlet Natalie, published via Pia. The union's vice-chair, the voice actor Nozawa Masako, did not attend and sent a video message in which she said she was disappointed that performers' voices were being used without permission, the same account says. Asked about the rule for deceased performers, the union's outside lawyer Ōkubo Shūichi said that making new works with a dead performer's voice takes work from performers who are active now, and that it is impossible to confirm whether the dead would have wanted the part, Natalie reported.

The union's vice-chair Ikemizu Michihiro said at the conference that building real cases of the guidelines in use, with partners such as the Itochu group, would lead to legislation, ITmedia reported. The announcement follows a court ruling: on September 30, 2026 the Tokyo District Court held that a voice can fall under publicity rights, in voice actor Tsuda Kenjirō's suit against TikTok's operator, though it dismissed his deletion request, as reported by ITmedia and in the union's own statement carried by Oricon. The guidelines also cite a resolution passed unanimously at the November 2025 world congress of the International Federation of Actors, which names informed consent, transparency and fair pay as global principles, according to the PDF.

For creators who use voice-cloning or AI dubbing tools, the guidelines change contract expectations rather than law. A studio or platform that wants to use a Japanese performer's voice can expect a request for stage-by-stage consent, a term of a year or less and separate payments, if the union's members and their agencies write the document into contracts. Nothing in it stops a developer outside Japan, or one that does not deal with union members, from scraping a voice today.

The union says it will keep pressing the government for a voice-rights law and will revise the document as partners add application cases; version 1.0 is open to change.