One evening in early September, T, a 28-year-old actor who asked to be identified by his first initial, took his seat in a rented Hollywood studio space in front of three cameras, a director, and a producer for a somewhat unusual gig.

The 2-hour shoot produced footage that was not meant to be viewed by the general public—not less than, not a human public. 

Quite, T’s voice, face, movements, and expressions can be fed into an AI database “to higher understand and express human emotions.” That database would then help train “virtual avatars” for Meta, in addition to algorithms for a London-based emotion AI company called Realeyes. (Realeyes was running the project; participants only learned about Meta’s involvement once they arrived on site.)

The “emotion study” ran from July through September, specifically recruiting actors. The project coincided with Hollywood’s historic dual strikes by the Writers Guild of America and the Screen Actors Guild (SAG-AFTRA). With the industry at a standstill, the larger-than-usual variety of out-of-work actors can have been a boon for Meta and Realeyes: here was a brand new pool of “trainers”—and data points—perfectly suited to teaching their AI to seem more human. 

For actors like T, it was an excellent opportunity too: a technique to make good, easy money on the side, without having to cross the picket line. 

“That is fully a research-based project,” the job posting said. It offered $150 per hour for not less than two hours of labor, and asserted that “your individual likeness won’t be used for any industrial purposes.”  

The actors can have assumed this meant that their faces and performances wouldn’t turn up in a TV show or movie, however the broad nature of what they signed makes it not possible to know the complete implications needless to say. Actually, so as to participate, they’d to sign away certain rights “in perpetuity” for technologies and use cases that will not yet exist. 

And while the job posting insisted that the project “doesn’t qualify as struck work” (that’s, work produced by employers against whom the union is striking), it nevertheless speaks to a few of the strike’s core issues: how actors’ likenesses may be used, how actors must be compensated for that use, and what informed consent should appear to be within the age of AI. 

“This isn’t a contract battle between a union and an organization,” said Duncan Crabtree-Ireland, SAG-AFTRA’s chief negotiator, at a panel on AI in entertainment at San Diego Comic-Con this summer. “It’s existential.”

Many actors across the industry, particularly background actors (also generally known as extras), worry that AI—very like the models described within the emotion study—could possibly be used to exchange them, whether or not their exact faces are copied. And on this case, by providing the facial expressions that can teach AI to seem more human, study participants may in truth have been those inadvertently training their very own potential replacements. 

“Our studies don’t have anything to do with the strike,” Max Kalehoff, Realeyes’s vice chairman for growth and marketing, saidin an email. “The overwhelming majority of our work is in evaluating the effectiveness of promoting for clients—which has nothing to do with actors and the entertainment industry except to gauge audience response.” The timing, he added, was “an unlucky coincidence.” Meta didn’t reply to multiple requests for comment.

Given how technological advancements so often construct upon each other, not to say how quickly the sphere of artificial intelligence is evolving, experts indicate that there’s only a lot these firms can truly promise. 

Along with the job posting, MIT Technology Review has obtained and reviewed a replica of the information license agreement, and its potential implications are indeed vast. To place it bluntly: whether the actors who participated knew it or not, for as little as $300, they seem to have authorized Realeyes, Meta, and other parties of the 2 firms’ selecting to access and use not only their faces but in addition their expressions, and anything derived from them, almost nevertheless and at any time when they need—so long as they don’t reproduce any individual likenesses. 

Some actors, like Jessica, who asked to be identified by just her first name, felt there was something “exploitative” in regards to the project—each within the financial incentives for out-of-work actors and within the fight over AI and using an actor’s image. 

Jessica, a Latest York–based background actor, says she has seen a growing variety of listings for AI jobs over the past few years. “There aren’t really clear rules straight away,” she says, “so I don’t know. Perhaps … their intention [is] to get these images before the union signs a contract and sets them.”

All this leaves actors, struggling after three months of limited to no work, primed to simply accept the terms from Realeyes and Meta—and, intentionally or not, to affect all actors, whether or not they personally select to interact with AI. 

“It’s hurt now or hurt later,” says Maurice Compte, an actor and SAG-AFTRA member who has had principal roles on shows like and . After reviewing the job posting, he couldn’t help but see nefarious intent. Yes, he said, in fact it’s helpful to have work, but he sees it as helpful “in the way in which that the Native Americans did once they took blankets from white settlers,” adding: “They were getting blankets out of it in a time of cold.”  

Humans as data 

Artificial intelligence is powered by data, and data, in turn, is provided by humans. 

It’s human labor that prepares, cleans, and annotates data to make it more comprehensible to machines; as MIT Technology Review has reported, for instance, robot vacuums know to avoid running over dog poop because human data labelers have first clicked through and identified thousands and thousands of images of pet waste—and other objects—inside homes. 

In terms of facial recognition, other biometric evaluation, or generative AI models that aim to generate humans or human-like avatars, it’s human faces, movements, and voices that function the information. 

Initially, these models were powered by data scraped off the web—including, on several occasions, private surveillance camera footage that was shared or sold without the knowledge of anyone being captured.

But as the necessity for higher-quality data has grown, alongside concerns about whether data is collected ethically and with proper consent, tech firms have progressed from “scraping data from publicly available sources” to “constructing data sets with professionals,” explains Julian Posada, an assistant professor at Yale University who studies platforms and labor. Or, on the very least, “with individuals who have been recruited, compensated, [and] signed [consent] forms.”

But the necessity for human data, especially within the entertainment industry, runs up against a big concern in Hollywood: publicity rights, or “the correct to regulate your use of your name and likeness,” in keeping with Corynne McSherry, the legal director of the Electronic Frontier Foundation (EFF), a digital rights group.

This was a difficulty long before AI, but AI has amplified the priority. Generative AI particularly makes it easy to create realistic replicas of anyone by training algorithms on existing data, like photos and videos of the person. The more data that is out there, the simpler it’s to create a practical image. This has a very large effect on performers. 

Some actors have been in a position to monetize the characteristics that make them unique. James Earl Jones, the voice of Darth Vader, signed off on using archived recordings of his voice in order that AI could proceed to generate it for future movies. Meanwhile, de-aging AI has allowed Harrison Ford, Tom Hanks, and Robin Wright to portray younger versions of themselves on screen. Metaphysic AI, the corporate behind the de-aging technology, recently signed a take care of Creative Artists Agency to place generative AI to make use of for its artists. 

But many deepfakes, or images of pretend events created with deep-learning AI, are generated without consent. Earlier this month, Hanks posted on Instagram that an ad purporting to point out him promoting a dental plan was not actually him. 

The AI landscape is different for noncelebrities. Background actors are increasingly being asked to undergo digital body scans on set, where they’ve little power to ward off and even get clarity on how those scans will probably be utilized in the long run. Studios say that scans are used primarily to reinforce crowd scenes, which they’ve been doing with other technology in postproduction for years—but in keeping with SAG representatives, once the studios have captured actors’ likenesses, they reserve the rights to make use of them ceaselessly. (There have already been multiple reports from voice actors that their voices have appeared in video games aside from those they were hired for.)

Within the case of the Realeyes and Meta study, it may be “study data” reasonably than body scans, but actors are coping with the identical uncertainty as to how else their digital likenesses could sooner or later be used.

Teaching AI to seem more human

At $150 per hour, the Realeyes study paid excess of the roughly $200 every day rate in the present Screen Actors Guild contract (nonunion jobs pay even less). 

This made the gig a gorgeous proposition for young actors like T, just starting out in Hollywood—a notoriously difficult environment even had he not arrived just before the SAG-AFTRA strike began. (T has not worked enough union jobs to officially join the union, though he hopes to sooner or later.) 

Actually, even greater than a normal acting job, T described performing for Realeyes as “like an acting workshop where … you get a likelihood to work in your acting chops, which I believed helped me somewhat bit.”

For 2 hours, T responded to prompts like “Tell us something that makes you indignant,” “Share a tragic story,” or “Do a scary scene where you’re scared,” improvising an appropriate story or scene for every one. He believes it’s that improvisation requirement that explains why Realeyes and Meta were specifically recruiting actors. 

Along with wanting the pay, T participated within the study because, as he understood it, nobody would see the outcomes publicly. Quite, it was research for Meta, as he learned when he arrived on the studio space and signed a knowledge license agreement with the corporate that he only skimmed through. It was the primary he’d heard that Meta was even connected with the project. (He had previously signed a separate contract with Realeyes covering the terms of the job.) 

The information license agreement says that Realeyes is the only owner of the information and has full rights to “license, distribute, reproduce, modify, or otherwise create and use derivative works” generated from it, “irrevocably and in all formats and media existing now or in the long run.” 

This sort of legalese may be hard to parse, particularly when it deals with technology that’s changing at such a rapid pace. But what it essentially means is that “you could be freely giving stuff you didn’t realize … because those things didn’t exist yet,” says Emily Poler, a litigator who represents clients in disputes on the intersection of media, technology, and mental property.

“If I used to be a lawyer for an actor here, I’d definitely be looking into whether one can knowingly waive rights where things don’t even exist yet,” she adds. 

As Jessica argues, “Once they’ve your image, they will use it at any time when and nevertheless.” She thinks that actors’ likenesses could possibly be utilized in the identical way that other artists’ works, like paintings, songs, and poetry, have been used to coach generative AI, and she or he worries that the AI could just “create a composite that appears ‘human,’ like believable as human,” but “it wouldn’t be recognizable as you, so you possibly can’t potentially sue them”—even when that AI-generated human was based on you. 

This feels especially plausible to Jessica given her experience as an Asian-American background actor in an industry where representation often amounts to being the token minority. Now, she fears, anyone who hires actors could “recruit a couple of Asian people” and scan them to create “an Asian avatar” that they might use as an alternative of “hiring certainly one of you to be in a industrial.” 

It’s not only images that actors must be fearful about, says Adam Harvey, an applied researcher who focuses on computer vision, privacy, and surveillance and is certainly one of the co-creators of Exposing.AI, which catalogues the information sets used to coach facial recognition systems. 

What constitutes “likeness,” he says, is changing. While the word is now understood primarily to mean a photographic likeness, musicians are difficult that definition to incorporate vocal likenesses. Eventually, he believes, “it should also … be challenged on the emotional frontier”—that’s, actors could argue that their microexpressions are unique and must be protected. 

Realeyes’s Kalehoff didn’t say what specifically the corporate can be using the study results for, though he elaborated in an email that there could possibly be “a wide range of use cases, similar to constructing higher digital media experiences, in medical diagnoses (i.e. skin/muscle conditions), safety alertness detection, or robotic tools to support medical disorders related to recognition of facial expressions (like autism).”

When asked how Realeyes defined “likeness,” he replied that the corporate used that term—in addition to “industrial,” one other word for which there are assumed but no universally agreed-upon definitions—in a way that’s “the identical for us as [a] general business.” He added, “We should not have a selected definition different from standard usage.”  

But for T, and for other actors, “industrial” would typically mean appearing in some type of commercial or a TV spot—“something,” T says, “that’s directly sold to the buyer.” 

Outside of the narrow understanding within the entertainment industry, the EFF’s McSherry questions what the corporate means: “It’s a industrial company doing industrial things.”

Kalehoff also said, “If a client would ask us to make use of such images [from the study], we’d insist on 100% consent, fair pay for participants, and transparency. Nevertheless, that shouldn’t be our work or what we do.” 

Yet this statement doesn’t align with the language of the information license agreement, which stipulates that while Realeyes is the owner of the mental property stemming from the study data, Meta and “Meta parties acting on behalf of Meta” have broad rights to the information—including the rights to share and sell it. Which means, ultimately, the way it’s used could also be out of Realeyes’s hands. 

As explained within the agreement, the rights of Meta and parties acting on its behalf also include: 

• Asserting certain rights to the participants’ identities (“identifying or recognizing you … creating a novel template of your face and/or voice … and/or protecting against impersonation and identity misuse”)
• Allowing other researchers to conduct future research, using the study data nevertheless they see fit (“conducting future research studies and activities … in collaboration with third party researchers, who may further use the Study Data beyond the control of Meta”)
• Creating derivative works from the study data for any sort of use at any time (“using, distributing, reproducing, publicly performing, publicly displaying, disclosing, and modifying or otherwise creating derivative works from the Study Data, worldwide, irrevocably and in perpetuity, and in all formats and media existing now or in the long run”)

The one limit on use was that Meta and parties would “not use Study Data to develop machine learning models that generate your specific face or voice in any Meta product” (emphasis added). Still, the range of possible use cases—and users—is sweeping. And the agreement does little to quell actors’ specific anxieties that “down the road, that database is used to generate a piece and that work finally ends up seeming loads like [someone’s] performance,” as McSherry puts it.

Once I asked Kalehoff in regards to the apparent gap between his comments and the agreement, he denied any discrepancy: “We imagine there are not any contradictions in any agreements, and we stand by our commitment to actors as stated in all of our agreements to completely protect their image and their privacy.” Kalehoff declined to comment on Realeyes’s work with clients, or to substantiate that the study was in collaboration with Meta.

Meanwhile, Meta has been constructing  photorealistic 3D “Codec avatars,” which go far beyond the cartoonish images in Horizon Worlds and require human training data to perfect. CEO Mark Zuckerberg recently described these avatars on the favored podcast from AI researcher Lex Fridman as core to his vision of the long run—where physical, virtual, and augmented reality all coexist. He envisions the avatars “delivering a way of presence as when you’re there together, regardless of where you really are on the planet.”

Despite multiple requests for comment, Meta didn’t reply to any questions from MIT Technology Review, so we cannot confirm what it will use the information for, or who it means by “parties acting on its behalf.” 

Individual selection, collective impact 

Throughout the strikes by writers and actors, there was a palpable sense that Hollywood is charging right into a recent frontier that can shape how we—all of us—engage with artificial intelligence. Normally, that frontier is described on the subject of employees’ rights; the thought is that whatever happens here will affect employees in other industries who’re grappling with what AI will mean for their very own livelihoods. 

Already, the gains won by the Writers Guild have provided a model for learn how to regulate AI’s impact on creative work. The union’s recent contract with studios limits using AI in writers’ rooms and stipulates that only human authors may be credited on stories, which prevents studios from copyrighting AI-generated work and further serves as a serious disincentive to make use of AI to put in writing scripts. 

In early October, the actors’ union and the studios also returned to the bargaining table, hoping to offer similar guidance for actors. But talks quickly broke down because “it is obvious that the gap between the AMPTP [Alliance of Motion Picture and Television Producers] and SAG-AFTRA is just too great,” because the studio alliance put it in a press release. Generative AI—specifically, how and when background actors must be expected to consent to body scanning—was reportedly certainly one of the sticking points. 

Whatever final agreement they arrive to won’t forbid using AI by studios—that was never the purpose. Even the actors who took issue with the AI training projects have more nuanced views in regards to the use of the technology. “We’re not going to completely cut out AI,” acknowledges Compte, the actor. Quite, we “just have to seek out ways which are going to profit the larger picture… [It] is de facto about living wages.”

But a future agreement, which is specifically between the studios and SAG, won’t be applicable to tech firms conducting “research” projects, like Meta and Realeyes. Technological advances created for one purpose—perhaps people who come out of a “research” study—may even have broader applications, in film and beyond. 

“The likelihood that the technology that’s developed is barely used for that [audience engagement or Codec avatars] is vanishingly small. That’s not how it really works,” says the EFF’s McSherry. As an example, while the information agreement for the emotion study doesn’t explicitly mention using the outcomes for facial recognition AI, McSherry believes that they could possibly be used to enhance any sort of AI involving human faces or expressions.

(Besides, emotion detection algorithms are themselves controversial, whether or not they even work the way in which developers say they do. Will we actually need “our faces to be judged on a regular basis [based] on whatever products we’re ?” asks Posada, the Yale professor.)

This all makes consent for these broad research studies even trickier: there’s no way for a participant to opt in or out of specific use cases. T, for one, can be completely satisfied if his participation meant higher avatar options for virtual worlds, like those he uses along with his Oculus—though he isn’t agreeing to that specifically. 

But what are individual study participants—who might have the income—to do? What power do they really have in this example? And what power do people—even individuals who declined to participate—need to make sure that they are usually not affected? The choice to coach AI could also be a person one, however the impact shouldn’t be; it’s collective.

“Once they feed your image and … a certain quantity of individuals’s images, they will create an countless number of similar-looking people,” says Jessica. “It’s not infringing in your face, per se.” But perhaps that’s the purpose: “They’re using your image without … being held chargeable for it.”

T has considered the likelihood that, sooner or later, the research he has contributed to could very wellreplace actors. 

But not less than for now, it’s a hypothetical. 

“I’d be upset,” he acknowledges, “but at the identical time, if it wasn’t me doing it, they’d probably work out a distinct way—a sneakier way, without getting people’s consent.” Besides, T adds, “they paid rather well.”