Japan Actors Union Sets AI Voice Consent Guidelines for 2,600 Members
The union's new rules limit consent to one year in principle and require separate payment for AI training and synthesized-voice use.
The Japan Actors Union published "Guidelines for the Use of Voice-Generating AI" covering its roughly 2,600 voice actor and actor members, according to ITmedia. The document sets three principles for AI use of members' voices: prior consent, management of voice data, and payment of appropriate compensation.
Under the guidelines, blanket consent covering all future uses in a single contract is not accepted; consent is limited in principle to a maximum of one year. Operators planning to develop or train a new AI model must give at least 7 days' notice, while generating voice for a work from an existing model requires 48 hours' notice.
The rules also cover prompting. Reproducing a specific person's voice by entering their name, role or voice characteristics into a prompt falls under the guidelines even if the operator never trained on that voice actor's recording data. Compensation for AI-generated voice is set separately from existing appearance fees, split into AI training and development fees and synthesized-voice usage fees. Converting Japanese-recorded voice into other languages via AI for overseas distribution requires separate permission and additional payment.
Operators are required to manage voice data, prompts and AI models with measures such as encryption and usage records, and to delete the data and stop providing the AI model if the voice actor withdraws consent and the operator agrees. Reproducing a deceased voice actor's voice by AI is prohibited in principle, with exceptional use requiring strict permission from the bereaved family or rights manager. In anime and foreign film dubbing, performances by actual voice actors are to be prioritized over AI.
ITmedia reported the guidelines on October 16, 2026. According to the background cited, past performances and voice data published online have been collected without permission for AI training, and lines or songs the voice actor never spoke have been generated in their voice. Japan has no explicit law comprehensively protecting an individual's voice itself. The union plans to push for legislation.
Quick answers
What are the Japan Actors Union's guidelines for voice-generating AI?
The guidelines set three principles: prior consent, management of voice data, and payment of appropriate compensation. Consent is limited in principle to a maximum of one year, and operators must give 7 days' notice to train a new model or 48 hours to generate voice from an existing one.
How is compensation handled for AI-generated voice?
Compensation is set separately from existing appearance fees, including AI training and development fees and synthesized-voice usage fees. Converting Japanese-recorded voice into other languages via AI for overseas distribution requires separate permission and additional payment.
Are the guidelines legally binding?
The union published them as guidelines rather than law, and plans to push for legislation in the future.