The Pulse
Japan Actors Union Requires Consent at Every AI Voice Stage
Japan’s Actors Union published guidelines on Oct. 15 requiring prior permission for AI use of performers’ voices, with separate consent for learning, replica creation and project use. The rules also set notice periods, compensation principl
AI.info Team · Written with AI ·
Japan’s Actors Union announced its guidelines for AI-generated voices on Oct. 15, setting rules for how performers’ voices may be used in AI development and services. The Japan Actors Union, known as JAU, says the guidelines aim to protect performers while providing direction for producers and AI companies. The cooperative says about 2,600 actors and voice actors belong to the union.
Consent at each stage of AI use
The guidelines identify three stages: analysis and learning, including providing voice data or entering prompts; generation and construction of a digital replica or synthetic performer; and project use of generated audio, including output, editing, distribution or sale. Before AI use begins, companies must obtain the performer’s explicit written consent. The document calls for independent, individual consent for each of the three stages and rejects blanket advance agreements for broad or unspecified uses. It also requires new consent for uses beyond the original agreement, such as another project or medium.
JAU’s document sets notice periods according to the use. For new replica or synthetic-performer development, or new AI-model development or training, written notice is due at least seven days before data is provided, recording begins or the development process starts. For output or editing within a specific project using a model already approved, notice is due at least 48 hours before the AI use or session. The notice is to describe the project, purpose and duration, the technology, the content and the data involved.
Terms and additional payments
Consent and related agreements should generally last no longer than one year. A longer term is permitted when there is an objective, reasonable production need, such as a long game or animation project. The guidelines reject indefinite consent and call for renewed written agreement if use continues after the agreed term. They also say performers should not face adverse treatment in auditions, casting, future work or other conditions for refusing or withdrawing consent.
JAU calls for AI-related compensation separate from ordinary performance fees. Its categories include fees for providing voice data or developing a replica, charges for using generated voices in a project, additional fees for advertising and fees for AI dubbing or other multilingual use. The guidelines do not set a single rate. They say compensation should take account of factors such as the medium, countries of use, subscription and audience scale, and the performer’s career level.
Human performers and deceased performers
The guidelines say human performers should be prioritized in voice productions and call for consultation and agreement before AI voices replace performers’ work opportunities. AI use of a deceased performer’s voice is prohibited in principle. The document provides a limited exception for a sequel to a series in which the performer appeared, if concrete production had begun before the performer’s death, or for a project the performer had agreed to in life. The exception is subject to conditions, including limiting use to what cannot reasonably be completed by a human performer, obtaining prior written consent from heirs and the former agency, and paying the heirs. Use for a new project the performer had not agreed to is not allowed.
The rules also call for encryption, access controls and records of data sharing. If a performer withdraws consent and the performer and company agree, the company must delete the performer’s audio data and prompts within 30 days and promptly disable models trained or fine-tuned with those materials. Performers may request information about data and model use when they have reasonable grounds to suspect their voice has been reproduced without permission.
An industry guideline, not a law
The guidelines are an industry standard without legal force, according to ITmedia’s report on the announcement. JAU says Japan lacks a comprehensive law specifically protecting a person’s voice and says it plans to press for legislation while seeking adoption of the guidelines across the industry. The document proposes measures for violations, including requests to stop distribution or destroy models; those provisions are part of the union’s guideline.
At the Oct. 15 announcement, JAU vice-chairman and voice actor Michihiro Ikemizu responded to objections that the process would be demanding.
“今までが簡単すぎた” — Michihiro Ikemizu, JAU vice-chairman and voice actor.
The phrase means, “Things had been too easy until now,” according to ITmedia’s report. The guidelines set out the union’s proposed baseline; their practical reach will depend on whether producers, platforms and AI companies adopt them in contracts and workflows.