Japan Actors Union releases AI voice consent and compensation guidelines

At a Tokyo press conference on October 15, 2026, the Japan Actors Union unveiled a framework requiring consent and compensation for any unauthorized generative AI replication of voice performances. The organization, which represents roughly 2,600 actors and voice actors, published a 22-article document officially named “Guidelines for the Use of Generative AI Regarding Voice Ver. 1.0.” The rules apply to any audio produced with AI technology that copies actors’ voices, whether through voice synthesis, deep learning, or machine learning. Kotaku reports that these guidelines rank among the first such standards released in Japan addressing AI voice use.

The union’s rules mandate consent before any stage of generative AI use, from training models on an actor’s performance to publishing generated content, and require clear disclosure of how the voice will be used. Each consent agreement must span training, content generation, publication, and any subsequent generated material, with written contracts detailing payment terms. The framework specifies three distinct points where written consent is required: capturing the voice for training, constructing an AI model from it, and deploying the output in a project. In principle, individual consent is capped at one year, after which renewed permission is needed for continued use.

When generative AI replicates a performance, the union treats it as a new performance that triggers additional royalties or compensation for the actor. Royalty calculations depend on how extensively and in what manner the voice is used, with separate fees possible for development, advertising, and multilingual exploitation. Payment categories beyond the standard performance fee cover learning and development costs, generative AI usage fees, added charges for advertising, and multilingual usage fees. The framework bans comprehensive buyout contracts, declaring them invalid, so performers cannot sign away all future rights in a single lump-sum agreement.

For game localization and English dubs of Japanese titles, the guidelines establish specific guardrails. When generative AI translates a performance into another language, the output must preserve the artistic integrity of the original, and the result cannot be used overseas without permission. The framework states that AI should not replace Japanese-language voice performances in dubbed foreign productions as a matter of principle. If AI dubbing is permitted, consent must be secured from performers, rights holders, and audio producers, with individual written contracts specifying compensation and terms, and the rules broadly prohibit using AI as a substitute in foreign language dubbing.

The guidelines also regulate deceased actors’ voices. Generative AI may replicate a deceased actor’s voice only if permission was granted during their lifetime, and users must additionally secure consent from the family and affiliated agency while paying royalties. The framework bans the practice in principle, with exceptions only for ongoing works or pre-approved projects. Even then, the bereaved family and the former agency must give written consent and receive suitable compensation.

Limited use of a deceased performer’s voice may be permitted for projects agreed upon before death, requiring written consent from the bereaved family and the former agency along with suitable compensation.

Firms that use a voice without consent may be asked to remove the content, destroy the AI model, and pay sales retroactive to the day unauthorized use began. However, the guidelines are voluntary and carry no legal force, so these remedies depend on industry cooperation rather than court enforcement. Japan has no comprehensive law protecting the voice itself, and Article 91 of the Copyright Act, which covers recording and filming rights, does not adequately address AI learning and generation of voices. The union is pressing lawmakers to create voice-rights legislation, a process that a union vice chair characterized as potentially taking several years.

The legal landscape is beginning to shift. On September 30th, the Tokyo District Court ruled that a person’s voice can be protected by publicity rights, marking the first such acknowledgment in Japan. The ruling emerged from a lawsuit by voice actor Kenjiro Tsuda over AI-cloned TikTok videos, though the court rejected the request for deletion because the videos had already been taken down. Anime News Network reports that Tsuda filed the lawsuit at Tokyo District Court in November 2025.

TikTok contends the narration uses a “universal male voice” created by training AI on a friend’s voice and denies any confusion or publicity-rights infringement, according to Anime News Network. TikTok’s Japanese subsidiary stated it is responding appropriately to the lawsuit and will refrain from further comments. Tsuda’s lawyer, Kei Hirano, said that unchecked unauthorized AI production could hinder the voice acting industry and that a win could clarify illegality and encourage actors to exercise their rights.

The union’s push for guidelines comes against a backdrop of widespread AI adoption in Japanese game development. GamesIndustry.biz reports that research found over half of Japanese developers use generative AI in their work, primarily for visual assets, images, character art, story generation, in-game text, and programming support. Meanwhile, videos and audio featuring AI voices identical to those of voice actors continue to spread online.

Masako Nozawa, who serves as vice chair of the Japan Actors Union and has voiced Goku for 40 years, described unauthorized AI use of voices as disillusioning and no different from theft. Nozawa, who turns 90 on October 25th and has voiced Goku since Dragon Ball debuted on TV in 1986, sent a video message for the October 15th press conference due to a scheduling conflict. She remarked that AI’s progress is amazing, but she finds unauthorized use of voices disillusioning, and she hopes the rules are followed so humans and AI can happily coexist. Kotaku reports that in a separate interview, Nozawa told NHK she cannot forgive unauthorized use of her voice, calling it theft, and said an AI version of her voice she heard lacked heart; AnimeMojo.com reports that Nozawa received the government’s Person of Cultural Merit honor last year, becoming the first voice actor so named, and that she recently joined the cast of Ranma 1/2 Season 3.

Union board member Bin Shimada, who voices Broly and Tomozo in Chibi Maruko-chan, said he was stunned and then angered by an AI-generated cover of Ado’s Usseewa using his character’s voice. Shimada said words he never said and songs he never sang were uploaded and monetized without permission, which he cannot accept. He drew a line between private enjoyment and public monetized exploitation, saying posting clips of performers saying lines or singing songs they never recorded for followers and money is unacceptable.

Union Vice Chairman Michihiro Ikemizu said the industry needs rules so AI can be used properly, safely, and with peace of mind. He noted that legislation to protect voices could still take several years, framing the voluntary guidelines as an interim step while the union pushes for legal protections. The union will strive to establish the guidelines as industry-wide rules and encourage the government to legislate regulations protecting voice rights.

The 22-article document also calls for credit labels on AI-generated voices and lets actors ask for an audit that reveals usage logs. Performers who refuse or withdraw AI consent must not be disadvantaged in casting, auditions, or employment conditions, and full AI replacement projects need prior written approval from a new Voice AI transaction fairness committee, whose details and timing the committee will set later.

The Japan Actors Union’s effort comes as other countries move to regulate AI voice use. In the United States, SAG and AFTRA merged in 2012 into a union of over 150,000 members, and the 2026 TV/Theatrical Agreement includes consent provisions for digital replicas used in dubbing into other languages. GamesIndustry.biz reports that in 2023, SAG-AFTRA members voted to authorize a strike against several major publishers over AI and job security, and after an 11-month strike approved the 2025 Interactive Media Video Game Agreement with AI guardrails that includes consent and disclosure requirements for AI digital replica use and lets performers suspend consent for new material generation during a strike. UKIE and law firm Sheridans launched an AI Compliance and Accreditation Service on October 19, 2026, to help UK studios use AI in line with legal and responsible standards.

However, the American model is difficult to transfer directly to Japan because performers there are divided among unions and agencies with different interests, making unified bargaining harder. Note reports that Yoshimoto Kogyo has over 6,000 affiliated talents, Geidankyo has 68 regular member organizations representing approximately 90,000 performers as of 2018, and the Japan Entertainment Artists Association has over 50,000 freelancer members. The scale of these rival organizations helps explain why no single body can claim to represent a consensus of Japanese voice owners.

The Japan Entertainments Management Entrepreneurs’ Association and the Japan Voice Actors Association reportedly support the standards, and most voice actors in Japan belong to the union or these two associations. The Japan Actors Union issued a 2024 joint statement with agency organizations including the Japan Management Association and the Japan Voice Actors Association, while other bodies pursue different priorities.

Japan’s government is facing questions on how to deal with generative AI in response to cyberattacks and copyright infringement, as Kotaku notes. The industry ministry has begun considering revisions to the Unfair Competition Prevention Law to address unauthorized AI-generated voices and likeness, according to The Asahi Shimbun.

In October 2024, Aoni Production partnered with CoeFont to build licensed AI versions of actors’ voices in English, Chinese, and other languages, with Nozawa and Banjo Ginga among the first two names, as reported by AnimeMojo.com. That licensed arrangement covered voice assistants, robots, and navigation systems but explicitly excluded anime and dubbing, showing a separate consent-based path from the unauthorized replication targeted by the guidelines. In 2024, a group of Japanese voice actors formed an organization protesting AI voice misuse, citing the lack of legal copyright protections, according to Kotaku.

The Japan Actors Union said technological advancements will not stop, but the “living, breathing human expression” and “respect for rights” must always remain at the core of AI use. The union made this statement alongside its announcement that it will push for industry-wide adoption of the guidelines and government legislation protecting voice rights, framing the guidelines as part of a broader effort to balance technology with performer rights.