Japan Actors Union Sets AI Voice Rules: Consent at Every Stage

The Japan Actors Union has issued guidelines requiring prior consent at "every stage" of generative AI use on voice performances, including before a model trains on an actor's work. The framework, reported by Automaton, targets unauthorised replication of voice actors — a group heavily exposed to game localisation and dubbing workflows.

The core rule is simple. If generative AI replicates a performance, the output counts as a new performance. That triggers additional royalties or compensation for the actor involved. For studios, it means AI-generated voice lines are not free assets; they're licensed work.

What does this change for studios?

Three concrete requirements stand out:

• Training consent: Developers and AI companies must obtain an actor's permission before training models on their performances, not just before shipping the output.
• Deceased actors: AI voice replication is permitted only if the actor granted permission during their lifetime. Developers must also secure consent from the deceased's family and their affiliated agency or company, and pay royalties.
• Automatic translation: AI-generated translations into other languages must preserve the "artistic integrity" of the original performance, and cannot be deployed overseas without permission — a direct constraint on machine-localisation pipelines that Japanese publishers have increasingly trialled.

The union framed the stakes plainly. "The Japan Actors Union will strive to not just establish these guidelines as industry-wide rules, but also strengthen its efforts to encourage the government to legislate regulations that protect 'voice rights'," it said in a statement. "Technological advancements will not stop, but the 'living, breathing human expression' and 'respect for rights' must always remain at its core."

That second sentence signals the real ambition: these are guidelines today, but the union wants statutory backing tomorrow. Japan's game industry sits squarely in scope. Research last year showed that over half of Japanese developers already use generative AI in game development, primarily for visual assets, images, character art, story generation, in-game text and programming support. Voice has so far been a smaller slice of adoption; these rules aim to keep it that way, or at least keep it paid.

Parallel pressure from SAG-AFTRA and UKIE

The Japanese guidelines land in an already-shifting global regulatory environment for performers. In 2023, SAG-AFTRA members voted to authorise a strike against several major publishers to protect performers from AI and ensure job security.

After an 11-month strike, members approved the 2025 SAG-AFTRA Interactive Media Video Game Agreement. It includes consent and disclosure requirements for AI digital replica use, plus the ability for performers to suspend consent for generating new material during a strike — effectively a kill switch on replica rights when labour action resumes.

The UK is moving on the compliance side rather than the performer side. Yesterday, UKIE announced a partnership with law firm Sheridans to offer an AI Compliance and Accreditation Service, supporting UK studios in using AI lawfully and responsibly.

For studios operating across regions, the picture is now a patchwork: Japanese union guidelines on voice rights, a US collective agreement with enforceable consent mechanics, and UK self-regulatory accreditation. A studio localising a Japanese game with AI-translated voice lines could conceivably hit all three regimes at once.

The Japan Actors Union's push for legislation is the signal to watch: if the Japanese government takes up "voice rights" as a statutory category, consent-and-royalty obligations for AI replicas would shift from voluntary guidelines to legal requirements for every studio shipping in Japan.