Creative industries' relationship with generative AI is often discussed as an unresolved cultural debate. In practice, at least around performer likeness and voice specifically, there's now real, binding legal and contractual structure, and a real enforcement case that shows what happens when a studio doesn't follow it.
What the new union agreement actually requires
In June 2026, SAG-AFTRA members ratified a new TV/Theatrical Agreement that builds on earlier AI and digital-replica protections, further restricting use of synthetic performances and adding new protections against AI replacing members' work. Under current SAG-AFTRA terms, signatory studios must notify the union and bargain before a digital replica can replace covered work. And, notably, vague contract language attempting to authorize "any and all AI uses" of a performer's digital replica is treated as unenforceable. That last point directly targets a specific tactic (burying broad AI-rights grants in boilerplate contract language) that had been a real point of concern earlier in these negotiations.
What California state law adds on top
Separate from union contracts, California's AB 2602 requires explicit, informed, written consent before a studio can train a model on, or deploy a synthetic version of, an actor's voice, a legal requirement that applies regardless of a specific production's union status, adding a second, independent layer of protection beyond whatever a given contract says.
The real enforcement case that shows what happens when this is ignored
In May 2025, SAG-AFTRA filed an unfair labor practice charge with the National Labor Relations Board against Llama Productions after Fortnite rolled out an AI-generated Darth Vader voice without notifying the union or bargaining first. SAG-AFTRA's argument was direct: the AI voice had effectively replaced union-covered actors' work without the union being told or given the chance to negotiate, exactly the scenario the notification-and-bargaining requirement above is designed to prevent. This is a genuinely useful case study because it's not a hypothetical policy question. It's a live test of whether these protections actually get enforced when a well-resourced studio deploys AI voice technology at scale.
Why this connects directly to the consent infrastructure in voice cloning generally
This dispute sits alongside the platform-level consent verification we cover in our piece on AI voice cloning. ElevenLabs' consent captcha and similar technical safeguards address unauthorized cloning by outside actors, while SAG-AFTRA's agreements and AB 2602 address a different but related risk: a studio or platform holder using AI voice technology on a performer's own established character or likeness without the specific notification and bargaining process performers' unions have negotiated for.
Why "where studios draw the line" now has a more specific answer than industry sentiment
Rather than a vague, evolving cultural consensus, the actual boundary in unionized productions is now reasonably well-defined: notify the union, bargain before deployment, and don't rely on broad boilerplate AI-rights language to skip that process, with California law adding an independent consent requirement on top for voice specifically. The Fortnite/Llama Productions case shows this isn't just a paper requirement; it's actively being tested and enforced when studios don't follow it.
What to actually watch next
The outcome of the NLRB charge against Llama Productions is a genuinely significant signal to watch, a ruling that enforces the notification-and-bargaining requirement meaningfully strengthens these protections industry-wide; a weaker outcome would suggest the gap between contractual protection and practical enforcement remains real, the same gap we found in our coverage of NYC's AI bias audit law, where a well-designed requirement still faced documented enforcement problems in practice.