The Ghost in the Greenroom: How AI Is Rewriting Hollywood
In September of last year, a young actress named Tilly Norwood began making the rounds in Hollywood — a highlight reel, a tidy Instagram presence, and the sort of girl-next-door appeal agents have built careers on for a century. Several were reportedly circling. There was only one complication, and it was not a small one: Tilly Norwood does not exist. She is a composite generated by software, the product of a company that bills itself as the world's first AI talent studio. And the question her brief notoriety raised is the one gnawing at the whole industry: if a studio can manufacture a leading lady from a text prompt, what happens to the women — and the grips, colorists, and boom operators — who used to do that work for a living?
The actors' union, SAG-AFTRA, which represents roughly 160,000 performers, didn't bother with diplomatic hedging. "SAG-AFTRA believes creativity is, and should remain, human-centered," the guild declared, opposing "the replacement of human performers by synthetics." Working actors were blunter. Mara Wilson, who as a child starred in Matilda, asked about the real young women whose faces were presumably scraped to assemble the phantom: "You couldn't hire any of them?" The honest answer is that, increasingly, studios would rather not — and to see why, you have to hold three threads at once: unresolved law, incidents that read like dispatches from a future arriving early, and labor numbers that are frankly brutal.
The law is where the real fight is, and it's messier than either side would like. In June 2025, Disney and Universal — two companies not famous for cooperating on anything — jointly sued the image generator Midjourney, accusing it of running a "virtual vending machine" that spat out endless copies of Darth Vader, Shrek, and Homer Simpson on demand. The case rests on one load-bearing question: whether training a model on copyrighted work is permissible as fair use. Knock that out and the economics of the entire enterprise change overnight. Disney's legal chief, Horacio Gutierrez, put it plainly:
"We are bullish on the promise of AI technology... But piracy is piracy, and the fact that it's done by an AI company does not make it any less infringing." — Horacio Gutierrez, Disney Chief Legal Officer, statement to CNN, June 2025
Here the story takes its first turn, because the early legal weather has been contradictory. In March, Anthropic agreed to a $1.5 billion settlement over books used in training, the largest on record — yet in February the Supreme Court let stand the principle that purely AI-generated art can't be copyrighted at all, because copyright requires a human author. Sit with that irony: the same logic that protects human creators may leave a studio with no ownership over its fully synthetic output. Judges remain split, and the Disney–Midjourney case itself is grinding toward trial later this year. Whoever wins sets the terms of the next decade.
While the lawyers argue, the work itself is being quietly reorganized — sometimes in ways that are not mundane at all, but genuinely strange. Consider As Deep as the Grave, an indie drama that premiered its first footage at CinemaCon this spring. Its priest, Father Fintan, is played by Val Kilmer — the Top Gun and Tombstone star who died in April 2025. Kilmer was cast years before his death but was too ill with throat cancer to ever shoot a scene; the performance on screen was reconstructed entirely by generative AI, trained on old footage and voice recordings, with the cooperation of his estate and his daughter Mercedes. The reactions split along a fault line that runs through this entire debate. Some called it a fitting tribute; others insisted there is "nothing of Val Kilmer in it except the resemblance." The wrinkle that makes the case so hard to file neatly is consent: Kilmer himself had embraced the technology while alive, using an AI voice tool to restore his speech for Top Gun: Maverick after cancer took his natural one. His family insists the digital resurrection is what he wanted.
That ambiguity is exactly why the Kilmer case is more instructive than a simple horror story. It is not the studios stealing a likeness; it is a family offering one. And it sits next to a very different use of the same tools, where scale is what does the talking. Short-form mobile series have become a $1.3 billion business, and one company backed by Jeffrey Katzenberg can reportedly generate a thirty-minute show for as little as $3,000 — programming that, made the old way, ran up to $300,000 an episode. That hundredfold collapse does not stay theoretical for long. Defenders insist the technology is a brush, not a replacement for the painter; Netflix's Ted Sarandos called it "a creator's tool, not a creative tool." Whether that distinction survives a budget meeting is another matter — and when an unconsented deepfake of Brad Pitt and Tom Cruise circulated early this year, SAG-AFTRA and the studios it usually battles both condemned the "blatant infringement." When the union and the bosses agree, you're looking at a threat that frightens everyone.
Then the jobs, where the numbers aren't disputed even if their cause is. Per Bureau of Labor Statistics data, Los Angeles County's motion-picture industry employed about 142,000 people at the end of 2022; two years later, roughly 100,000 — a loss of 42,000 jobs, nearly a third of the workforce, in two years. Regional TV shoot days fell from a 2021 peak of 18,560 to 7,716 by 2024. But honesty requires the caveat that breaks the easy narrative: a spring study from Otis College of Art and Design, which found California's creative economy shed 114,000 jobs between 2022 and 2025, pointedly declined to pin the carnage on AI alone. The 2023 strikes froze production for the better part of a year; streaming consolidation gutted divisions; productions fled to Britain and Georgia chasing tax breaks LA couldn't match. The machine arrived to find the patient already weakened. Yet the shape of the loss is the tell — the eliminated roles cluster in repeatable mid-career work an algorithm can approximate, while postings for engineers who can wire generative tools into a VFX pipeline sit open past 120 days at software-industry salaries. The industry isn't just contracting; it's hollowing out in the middle and bidding up at the technical edge — the signature of a structural shift, not a cyclical dip.
It is tempting to file all this under the long history of Hollywood panics — the talkies that were going to end careers, the television that was going to empty the theaters, the CGI that was going to retire the stuntman. The industry has always swallowed its disruptors and gone on making money. But that comforting analogy is where the second twist hides, because the earlier shocks changed how films were made without changing who got to make them: a sound engineer replaced a title-card writer, a visual-effects house replaced a matte painter, and the total number of people drawing a paycheck kept climbing. What is different now is that the technology does not just shift the work from one set of hands to another. It can perform the task with no additional hands at all. A model that generates a thirty-minute episode for the price of a used car is not a new kind of camera; it is a substitute for the people who used to stand behind it. That is why the comparison to past manias only goes so far, and why the people living through this one are right to feel that the ground has moved in a way it did not before.
So where does the ghost actress finally resolve? Not in the courtroom or the spreadsheet, but in the negotiating room. The major guilds return to the table with the studios this year, contracts expiring at the end of June; the last round produced strikes of 118 and 148 days, and that was before any of this crested. AI guardrails and likeness control aren't side issues now — they're the table. The sharpest reframing came from inside the union. SAG-AFTRA president Sean Astin declined to play the doomed Luddite: "It's not the first shoe to drop," he told Variety. "We had a 118-day strike, fighting very hard to put AI provisions and protections in place... we're going to meet the challenge."
Because the machine is not a verdict; it's a tool, and tools take the morality of the hands that hold them. The same engine that fabricates a Tilly Norwood to avoid paying an actress can let a grieving family finish the role their father never got to play, or let one independent filmmaker make something she could never have afforded a crew for. Both futures are being built right now, in the same edit bays, by people who don't agree which one wins. That a new machine on the floor changes everything about how the work gets done, and nothing about the fact that someone still has to decide what the work is for, is an old truth the working person knows in his bones. The greenroom has a ghost in it now. Whether she replaces the living or works beside them is a question no algorithm will answer. We will.