How SAG-AFTRA union views AI and gaming negotiations | Sarah Elmaleh interview

Voice actor Sarah Elmaleh spoke at the Game Developers Conference last week, but she also had a side duty, as she is chair of a union committee for SAG-AFTRA, which is in negotiations with the game industry over the impact of AI on performers who give life to video game characters.

Elmaleh is a voice actor who has appeared in triple-A games including Helldivers 2, Hi-Fi Rush, Halo Infinite, Gears 5 and indies like Afterparty, Pyre and Gone Home. So she knows the contract being negotiated will directly affect her livelihood.

She’s also chair of the Interactive Media Agreement Negotiating Committee, which works closely with the lawyers representing the union in negotiations.

Elmaleh’s directing credits span production scope from Goodbye Volcano High and 2024 IGF Independent Games Festival Nominees The Wreck to Fortnite and Call of Duty. She is a leading voice in organizing, advising, and educating in the games industry and academia, and is instrumental in the current negotiations between SAG-AFTRA interactive performers and video game companies.

GamesBeat: What’s union life been like?

Sarah Elmaleh: It’s been busy. Very active. It takes a lot of my brain space.

GamesBeat: Can you talk about what you’re responsible for? Here at GDC, and at the union in general.

Elmaleh: I joined the committee a number of years ago. Then I was asked to serve as chair. As chair my job is to bring committee conversation, debate, discussion, both of waivers as a standing committee, special exceptions that productions might wish to ask for–as long as they don’t harm precedent and create work, we can approve those. That’s always an exciting part of this job.

Obviously the more prominent spotlighted part of this job, or this role – I say “job,” but it’s a volunteer position, not paid – is around leading negotiations for the interactive media agreement, the main agreement we have. Not the only one, but the main agreement we have that covers interactive work for SAG-AFTRA performer members.

It’s been a long negotiation. We’ve been negotiating this since 2022. As a result, we’ve cleared a lot of the issues on the table on both sides. It’s come down and focused on this final sticking point or final issue, which is AI protections. AI protections that don’t leave any performer out of the picture, so that everyone can have the transparency, consent, and compensation they need to participate in this, to have an equal space in building a paradigm where this tool is used ethically and they can benefit from that as a revenue stream. Details around our AI proposal is where we’re stuck currently.

GamesBeat: What is that conversation like as it goes back and forth? What are the issues that each side cares most about?

Elmaleh: We care most about those three principles. We would be willing to collaborate and tune the proposals further. We’ve put together as thorough a picture as we think might cover the intended use. Those three conditions–everything is built on the spirit of those three principles. Transparency, consent, and compensation. Transparency creates an opportunity for meaningful consent, not partial consent or misled consent. And then proper compensation. If value comes from your performance, it’s carried through when you create more material from that performance despite you not being present.

I would like to say that we were having conversations in response that would help us tune to the use that people want to use it for, but it hasn’t panned out that way. The nature of our negotiations is that we deliberate separately as our different groups, the bargaining group and our committee. What actually gets made is through our lawyers across the table. It hasn’t been–I wish I knew more about what folks wanted to use this for, because we could craft something that would be suited to that.

GamesBeat: I’m always in contact with startups in the space. I see what they want to do. I wonder, though, what the game companies would want to use this for, and also what’s going to be allowed. Different examples–the Replica thing was surprising to me in some ways, that there was an uproar over it. I don’t know if it wound up being as controversial as the headlines suggested. Is that an example of something that you’ve been able to figure out?

Elmaleh: For that, it was just the communication around it. Members are more activated around these issues and paying attention to union involvement around these issues than anything else, understandably. It’s an unprecedented threat, potentially. Folks who wanted that update were–the fact it was a surprise at all set off alarms for them. Institutionally we’ve extended ourselves. Okay, you’re paying attention. This is good. This is important. Let’s offer you the full agreement and be available to answer further questions.

Once those things were provided and the process of how we came to that agreement, the fact that our committee does this work in games–the work is close to our hearts. It’s close to our careers. The fact that it was us that deliberated on it and we were willing to show our work, you didn’t see the controversy continue to escalate after that point. That reassures me that the contract itself is protective and reasonable. I think folks were really activated around the surprise of it existing.

GamesBeat: I saw another similar case where they were going to modify voices to have different accents. What was interesting about that, it would provide an opportunity for people to get more work beyond the language in which they’re fluent. You could turn someone’s voice from, say, German-sounding to Irish-sounding, and create something that was not spoken or performed by the actor, but they would still get that work. That sounded like a scenario that could be favorable to the artist.

Elmaleh: There’s a lot of–actors take on a certain degree of risk, or operate without a certain amount of information in general. Especially in games. What you’re looking at with the use of this tool is a much wider and maybe more consequential risk window. The way you compensate for that is increased communication. That’s part of what the contract is built to encourage. Invite me into assessing that risk window with as much information as you can provide for me. Then that empowers me to make a meaningful decision to take that risk with you or not.

It’s also a calculated risk that we’re used to doing already. We don’t get much information about projects to begin with. We don’t get scripts in advance. There’s this moment in the booth where we’re presented with something, and right now we have that one moment to capture and say, “Oh, I don’t think I would say that, as an artist or someone from this background.” There’s a collaboration moment or a negotiation moment that allows us to take that risk and save something that could potentially have a negative effect, either on the game itself or my personal feeling about it. That’s the one piece of the current risk window that we lose when we start to use this tool. The contract is trying to recapture that conversation and make it a contractual obligation, as much as we can.

GamesBeat: Another one I remembered was the actor who died, and then they used that actor’s voice in a subsequent work, compensating the actor’s family for it. They had permission to do that, and to involve AI in making that happen. Does that also seem like a reasonable outcome?

Elmaleh: I’m pretty sure it’s in our proposal. The authorized representative, including someone’s estate, would be the person that’s in that consent pipeline. You’re still getting consent from someone who would be impacted by the use of the technology.

GamesBeat: Are there some scenarios that seem to be on a better side or a worse side?

Elmaleh: It’s sort of a shorthand for this, or a signifier of this–maybe a cheat sheet? But I feel like the folks who come to us to talk directly tend to be more on the good side. The folks who aren’t talking to us tend to be more on the–-it’s a good precondition. If you’re willing to come and ask actors what we need and agree to those terms, that’s a pretty good indicator that you’re trying to do this in good faith. If you’re trying to hide or evade that, maybe what you’re hiding is not so fantastic. That’s my big clue right now.

GamesBeat: Deep fakes, obviously there’s some bad stuff there. But there are some legit deep fakes as well. We’ve seen cases like Genies where a celebrity likeness can make a virtual appearance somewhere and generate revenue for someone who can’t make it there in person to voice a performance.

Elmaleh: The thing that I heard around what you just said–I think I maybe have to evolve this metaphor as I work on it and try it out. Something that occurs to me is there are big fears around full displacement. This disconnect or severing of your connection to your material and stuff that’s intimately from you as a performer, whether your voice or your likeness or any of these things. It’s really of your body and your soul. But when you have transparency, consent, and compensation – when you have this collaborative, fully aware dynamic that’s based on a degree of mutual respect – then you’re participating in this use in a meaningful way.

Instead of being displaced, it’s more like a prosthetic extension of yourself. It allows extra activity. You have an extra arm, essentially. The value comes back to you. It’s still connected to you. That interests me as a way to think about this that’s maybe a bit more reassuring. More work can be done. The value comes back to you. The consent comes from you. You’re not severed from your own value in that way.

GamesBeat: There’s another argument that the tech people would make, that they can also relieve an actor or performer of tedious work. If you have to do 100 different versions of a scream or something like that for a 30-hour game, maybe you just use AI to fill out things like that. Again, consensually, that sounds like a good use of technology.

Elmaleh: It’s certainly a barrier to existing game jobs. It’s a willingness to do what I would consider vocal stunt work. There’s a degree of physical toll and risk in doing that work. Folks who are not comfortable taking that risk on for whatever reason are not doing those jobs. If that appeals to them – if that role suddenly opens up where they can bring their soul and creativity to one aspect of this and have essentially a digital stunt double for vocally stressful work – that’s a use case that I’ve heard come from actors as something they would desire.

I love doing efforts. I love directing efforts. I’m Miss Collaboration City. I think there are character moments in all of these things. But as someone who chairs this committee, my job is to see all these affordances and create an ethical framework for all of them. It’s possible, and important in fact, to have those parallel conversations. This is what’s ethical. This is what’s possible. This is what creatively interests me. This is what creatively interests someone else, or doesn’t. Laying that groundwork so it supports all of these different decisions and conversations as much as we can. It’s a tall order, but that’s the task.

GamesBeat: We’re bringing up all these novel use cases, where new technology is coming out and being applied in creative ways. I can understand how, when you’re in the position you’re in, you can’t necessarily foresee all the innovations coming down the pipeline. What are you trying to do to futureproof the agreement, making sure that when new technology and new applications come out, you’re not coming back to the negotiating table to discuss every single new use case?

Elmaleh: Well, we’ll certainly be coming back every three years to discuss it. That’s just going to be–I think we’ve all accepted that the discussions around this technology will be ongoing for a while. My approach, or the way I see our approach, is to focus on not so much the details of the technology, or the terminology around it, or specific use cases, but to focus instead on the verb of creating material that derives from an actor without the actor present.

There are going to be a lot of different ways to do that. But we’re trying to say that if that’s what you’re trying to do, then here’s what you owe to the actor in the way of information, consent, and compensation. The compensation piece is probably the piece that we’ll see the most need to adapt around what development looks like, what the production pipelines look like. That I can see changing over time. But the basic principles of–you have to talk to your actors. Then whatever you want to do is driven back into that conversation. Fingers crossed that that’s futureproof. I don’t know, but that’s the idea.

GamesBeat: I’ve been bringing up hypotheticals and you’ve been talking about risks. But the question might be–if we look at the reality of what you’ve seen already happening, do you think we’re facing a lot of bad behavior already? Do you see a lot of good behavior? What’s the reality on the ground right now?

Elmaleh: The frustrating part to me right now is that there’s a lot of sell energy around the tech itself, a lot of top-down mandates around the use of this tool. But the conversations from developers who would actually be using it, implementing it, that’s much quieter. Maybe a little more shy? Actors, too, there’s a lot of loud fear and panic, which is completely understandable, and then maybe quieter interest or curiosity. There’s this desperate need for a grounded, practical, mutually respectful middle ground that’s devoid of utopia sell and also full-on panic.

When you get into the details around discussing these use cases and personal preferences and how you lay a groundwork where your comfort zone is relevant and important and meaningful, then you can start to clear some of the panic. On our committee I’m probably the more–I like my day at work. I love my day at work. I love directing this stuff. That’s the most important thing to me. I’m more allergic to this tech than some other folks. If you look at some recent surveys, there’s a split in the community, almost 50-50, from the National Association of Voice Actors around openness to this technology. I’m glad and grateful that our committee represents the spread of that interest and curiosity and tolerance for this.

Having conversations, good faith conversations, with people like Replica and others who’ve come to talk to us respectfully, that’s cleared a lot of my concerns. Realistically, it would come down to, do I trust this individual developer? Do I trust this individual use case? What can you show me about your guardrails and your scope? That would reassure me. I would take it probably case by case. That’s where I’m sitting right now.

It’s very difficult to see what’s out there on the ground in terms of use, because there are exploitative uses and all kinds of things running rampant. There’s a lot of energy and excitement around tools, but it’s less clear to me what developers are actually interested in. That’s why it’s been so exciting to come to GDC this week, because I can have some of those honest conversations in a safe space and start to get a picture. I hope that those continue.

GamesBeat: For the AI and games space, have you made the rounds and talked to many companies so far? The ones that are specifically in AI as it relates to games.

Elmaleh: I probably can’t speak to specific conversations until these things become public. I will say that our door is open. We want to talk to people. Anyone who wants to talk to us and ask how they can treat their actors ethically and collaborate with SAG-AFTRA–if that matters to them, we want to talk. Like I said, we’re open-minded. We want to do this ethically and create a paradigm for everyone, regardless of their interest or comfort with this tech. No one is barred from our door.

GamesBeat: Obviously you’re mostly working with professional actors at the highest levels. The games space–there’s always been an independent scene, but we’re seeing more and more examples of high quality indie games reaching mass market appeal. I can understand how some of the protections would be less prohibitive at larger developers than they would be at smaller developers. How flexible is the framework to be working between developers of different sizes and different budgetary restrictions on the ways they can deploy actors?

Elmaleh: I’m glad you mentioned indies, because we have a new tiered agreement. That’s been one of our biggest goals and focused efforts this week, around evangelizing that, spreading awareness of our tiered agreement. It’s the expansion of our original low budget agreement, which was designed to cater to indies and smaller budgets. You can find it all outlined on our website. It’s for discounted rates for budgets in a couple of different tiers, from $250,000 to $15 million. The prior budget cap was $1.5 million. We’ve been striving to stay in tune with scope and funding and budgets as they evolve. Obviously it’s changed since the number of years ago when we did the original. We have those protections in there too, the kind we’re bargaining for on the main agreement.

As someone who started out doing mods, working with folks of various budgets and raising my rates and then eventually making them union par, eventually becoming union, the idea that it’s too expensive to collaborate with actors at some threshold kind of confuses me. I’ve always collaborated with developers at every budget threshold. If you’re willing to–the barrier there actually feels, to me, a bit like it’s just your willingness to collaborate with people. Actors are easy to find. We all want to work.

I question the idea that there will be indies who have to use AI in order to make their games. I understand that budgets are incredibly real. Accommodating them is important to me. Understanding where they’re at is important to me. That’s why our waiver process is there. That’s why we’ll continue to take feedback on this agreement and all of that. But I think it’s really a question of what value you place on performance in your game. What value do you place on collaboration? How grounded is it in your gameplay experience? What are your values, ethics, interests? What’s the story you’re trying to tell? There’s a whole host of other questions. It’s not as simple as, “I need to use AI because it’s cheap.” There are some further questions I would have about the nature of the project.

Shannon Woodward, who was at our mixer this week–you might know her as Dina from The Last of Us. She was there promoting this agreement. She wants to work with indies. A lot of us do. I started in indies. As I started doing more triple-A, I never wanted to leave indies behind. I just felt that there are aspects of a union contract that really matter to me around pension and health contributions, or in this case protections and enforceability around something like AI, when this technology suddenly comes on the scene. It speaks to both established actors who love the heart-forward storytelling that’s present in indies, the performance-forward storytelling that’s in indies, the diversity of storytelling in indies–that’s really artistically motivating to folks, even if indies might think that they don’t have access to them based on their profile.

It’s equally true that this next generation of actors coming up into this space may want to seek the protections of this contract. Again, the consequences of–what’s the worst that could happen if you just invoice for your work? Which is often how a lot of indie voice-over works. I don’t have a contract, but what’s the worst that could happen? The worst that could happen right now is quite bad. Moving into a space where more of this work is covered by the union at all different levels of budget and experience–that’s a large task, and it involves a lot of advocacy and education work on our part, but I think that’s the safest paradigm for everyone moving forward. That’s my personal feeling.

GamesBeat: We’ve had situations like where OpenAI’s CEO said she didn’t know whether YouTube videos were used in training their models. Is there a viewpoint as well on the notion that–maybe they might want to be transparent, if you take them at their word, but they just don’t know the answer as to whether or not your stuff is being used in their models. Maybe it never actually spits out your stuff, but is it okay to have your stuff in the model, informing this thing on what it spits out? Transparency might not be possible when you see the output.

Elmaleh: It’s interesting to talk to folks who’ve been in procedural generation for a long time, and other forms of AI. There’s such a major difference there. There’s combinatorial, exponential possibility here, but the basic set of the material you work with is more transparent and knowable. Then you move into this space where we’re losing our ability to track what’s inside this thing. That should be scrutinized, I think.

GamesBeat: You have the New York Times suing about that sort of thing.

Elmaleh: Exactly. That piece of what’s allowed to be in there is being worked out and talked about and addressed in legislation with a number of different cases and bills. If you want more information about that, I have some names of acts in my head, but there are other folks like Duncan Crabtree-Ireland who are very involved in the legislative side. A number of folks at SAG-AFTRA are really tracing those bills that have to do with ingestion, if you’re curious.

GamesBeat: Have you done surveys of your membership around AI and what they think about the different things we’ve brought up?

Elmaleh: The process we have is more–we have wages and working conditions meetings. That’s how our proposals are put together. It’s how membership participates directly in what proposals are put on the table. We heard a range of massive concern. Other folks who were maybe quieter, there was interest. We’ve relied on that process, which is a standard process for all negotiations at our union, as well as individual conversations and temperature checks.

I found NAVA’s poll very compelling, because it cuts across a lot of union status in general. It’s a snapshot of voice actors in a larger sense. We don’t only have voice actors on this contract, just to be clear. We also have movement performers, folks who do performance capture. All aspects of performance are increasingly involved in making games. But they ran an actual poll, and I found that pretty compelling.

GamesBeat: With the number of game to film and TV adaptations on the rise, especially ones that are resonating with audiences outside of gaming–I don’t know if that has sparked more interest from actors in exploring roles within games. Is that something where, because it’s more prominent, people are taking it more seriously? Is it raising the profile of your advocacy efforts?

Elmaleh: I think there’s a realm of–Shannon Woodward is an interesting example to me as someone who has a storied, long on-camera career, but loves games. She’s done a few games and is really passionate about increasing work for herself and other actors who have more experience in other mediums coming to games. We’re also in a space where–when I first started coming to GDC, I was one of only a few voice actors here. I’d always wanted to do games, but I started in something else and kind of flopped sideways into it.

Both of those things, I think, are true. There’s a world of on-camera actors who maybe love games, but haven’t thought about or had access to the work itself. They might be curious about it, but it might be slightly invisible, and we should reach them. As well, more and more generations of actors grew up playing these things and loving them and seek to specialize in them from day one.

The commonalities, I would say, derive from the fact that we share–Zeke Alton, who’s spent the most time immersing himself in the technology and the use cases as they evolve, was relied on in the TV and theatrical negotiation for their proposals. He’s part of our committee as well. We both have a shared bedrock. An institutional position on these things is beginning to emerge. At the same time I’m very sensitive to the nuances of this work. We’ve been very insistent on those as well, where things are different and where special accommodations need to be made.

We’re all one acting community at this point. To your point, it’s increasingly evident, and as we near this question of whether or not we’re going to be on strike, people should expect solidarity across different disciplines and visibilities. The parallel is, 20 years ago, if you did television, you did television. Now film directors want to do television. TV directors are doing film. There’s a whole theater piece. It’s media. People are crossing the lines. We’re going to see more people crossing into this very successful side of entertainment.

GamesBeat: Was there much discussion about unions more broadly here at GDC?

Elmaleh: It seemed like it. It’s encouraging for me because I care about developers. I don’t think this is a controversial position for me to take given my role, but I think unions are an especially effective tool in acting as a counter-force to forces of change and volatility that aren’t geared toward the individual sustainability of a career in this space. I feel very strongly, not only from a human interest point of view and just caring about human beings, but that the quality of games suffers. We lose games when we lose people, when we burn them out and create such an unstable ecosystem that they have to seek more sustainable paths.

I’m excited for devs that there’s an interest in this, that there’s awareness, that information is being shared, that support is being offered by unions themselves. I’ve offered myself as a resource and offered myself in solidarity to developers on that front forever, and I’ll continue to do so.

GamesBeat: Do you think you’re heading toward a strike? Is there any characterization that you’ve communicated on that front?

Elmaleh: We’re certainly at a standstill. If no more movement comes our way then we’ll have to act in order to create that movement. We did see some movement created by the strike authorization itself. If it takes a strike in order to get us across the finish line and make sure we protect all performers in the use of this technology, that’s what we’ll have to do.