Hollywood’s generative-AI debate has moved beyond theoretical arguments about whether the technology will arrive. It is already being used across production, audio, development, marketing and experimentation, while studios and technology companies race to define the terms under which it is used. What remains deeply unsettled is the part that matters most: who controls the tools, the performances, the intellectual property and the creative decisions.
The current picture is not one unified “AI Hollywood.” It is a patchwork of very different bets. Netflix has bought a filmmaking-focused AI company and says roughly 300 programs in its library have used generative AI in some capacity. Amazon MGM has created an AI Studios leadership role while stressing that use is not mandatory. Lionsgate has put a chief AI officer in place and expanded its relationship with Runway. Google DeepMind is pursuing partnerships with filmmakers and venues. Meanwhile, startups are promoting fully AI-generated films, synthetic performers and creator-facing video platforms.
That is less a single production revolution than an arms race for knowledge. Each company wants to learn where AI is actually useful before its competitors do, while avoiding a public backlash, labor conflict or rights controversy severe enough to poison the experiment.
What “generative AI” means in this context
Generative AI refers to systems that create new output from prompts or other inputs: images, video, audio, text and related material. In film and television, that could mean concept images, altered shots, virtual production material, animation, voices, storyboards, audience analysis or short-form extensions of existing properties. It does not describe one consistent workflow or a fixed quality level.
That distinction is important. A production that uses an AI-assisted process at one stage is fundamentally different from a project that is presented as fully AI-generated. Yet the public response can flatten those differences, especially when viewers feel a finished work looks artificial, when credits do not clearly explain its use, or when a recognizable person’s voice or likeness is involved.
There is also a business distinction hiding inside the technical language. Using an AI tool to speed up a task is an efficiency play. Using data and models to identify a neglected audience for an older film, create a new format around owned intellectual property, or make an immersive venue experience is a search for new revenue and distribution opportunities. Studios are pursuing both.
Netflix’s large bet makes AI use harder to dismiss as a fringe experiment
Netflix’s purchase of Ben Affleck’s InterPositive in March, reported at $587 million and widely characterized as a deal approaching $600 million, gave the streaming giant a particularly visible position in the race. The acquisition put a tools-focused company with filmmaking roots inside a major entertainment platform, signaling that Netflix sees AI as a creative-production issue rather than solely a back-office technology initiative.
Netflix has since said that about 300 programs have incorporated generative AI in some form. That headline is meaningful, but it should not be treated as proof that hundreds of series or films were generated by machines. “Incorporated” can cover a wide range of uses; the available information does not provide a project-by-project breakdown. The more defensible takeaway is that AI-assisted work is already broad enough within Netflix’s operation that the company wanted to disclose its presence while emphasizing filmmaker-led production.
The streamer is also experimenting with AI animation, although it has said no AI-animated projects are actively in development. Separately, it has worked with audio company ElevenLabs on an AI replica of Gene Wilder’s voice for the upcoming reality program Wonka’s The Golden Ticket. That example goes straight to the most sensitive fault line in the entire conversation: a synthetic voice is not just a production asset. It can invoke identity, legacy, permission and the boundaries of a performer’s work.
For audiences, disclosure is likely to matter as much as the tool itself. The public may accept a technology differently when it supports a creator’s specific vision than when it appears to imitate a human performer or conceal how familiar material was made.
Amazon MGM and Lionsgate frame AI as optional infrastructure and library strategy
Amazon MGM appointed Albert Cheng as its first head of AI Studios, assigning him to integrate AI across the production pipeline without mandating its use. That wording is consequential. A mandate suggests a replacement-oriented directive; optional infrastructure presents AI as a resource creators may choose when it serves a project.
Amazon MGM has also used its “AI on the Lot” event to discuss the technology’s role and announced AI-animated shows through its GenAI Creators’ Fund. One project associated with The Book of Life director Jorge R. Gutierrez did not proceed after he withdrew amid criticism of the planned AI-animated work. The episode demonstrates the central practical constraint: an initiative can have studio backing and still be reshaped by artists and audience reaction.
Cheng’s stated pitch is reduced time between seasons and more capacity for creators to make series and films. That is an attractive proposition in an industry where long waits between installments can test viewer loyalty. But speed is not a creative outcome by itself. Faster pipelines still require scripts, performances, visual judgment, legal clearance and human decision-making if the final work is to feel intentional rather than merely plentiful.
Lionsgate’s approach is notably tied to the value of a catalog. The studio hired Kathleen Grace as its first chief AI officer and has an existing arrangement to train a model using the company’s film library to generate new content. In June, Lionsgate expanded its Runway relationship through an equity stake intended to support an AI-generated short-form series based on one of its properties.
Grace has described the broader opportunity as finding audiences and new ways to distribute material a studio already owns, rather than treating AI only as a cost-cutting mechanism. Large language models, the systems often used to analyze and generate text, could theoretically help identify communities interested in particular catalog titles from large pools of social data. But identifying an audience is not the same as earning that audience’s affection. Marketing intelligence can suggest a niche; it cannot guarantee that an AI-derived extension of a beloved film will feel worthwhile.
Google DeepMind is seeking creative legitimacy through partnerships
Google occupies an unusually complicated position in entertainment. Its AI training practices have generated concern among people working in media and creative fields, but its DeepMind unit is also trying to build projects with filmmakers and other specialists. The strategy appears to be collaboration rather than simply presenting a general-purpose model and asking Hollywood to adapt around it.
DeepMind has been involved in the AI-powered production of The Wizard of Oz at Sphere in Las Vegas, and it may be involved with the venue’s forthcoming The Rocky Horror Picture Show production. It has also worked with directors including Darren Aronofsky, whose Primordial Soup partnered with Google, and Doug Liman, who helped develop the immersive film Asteroid for an extended-reality app.
DeepMind’s $75 million investment in A24, alongside research into tools with the studio, triggered a backlash precisely because A24 carries a reputation for filmmaker-driven and independent-minded work. Its response—that participation is preferable to sitting on the sidelines—captures a growing view among some creators: if these systems are going to affect filmmaking, artists should have influence over their development rather than refusing every engagement.
That is not the same as an endorsement of every use case. It is an argument for having a say in how the tools are shaped, what safeguards are built and what permissions govern the material they touch.
The startups show both the ambition and the controversy
Newer companies are frequently more willing to test the limits of what audiences will accept. Particle 6 drew sharp criticism around Tilly Norwood, its synthetic “actor,” after suggestions that the character could be signed to an agency. The reaction fed fears that digital performers could displace human actors and contributed to new SAG-AFTRA guidelines concerning synthetic performers.
Particle 6 chief executive Eline Van der Velden argues that Tilly is intended for underserved or niche audiences and not as a human replacement. The company’s Tilly-led film Misaligned began an expected 12-month production cycle this month. Van der Velden has also compared the creative direction of the character’s performance to puppeteering and motion capture, suggesting that performers could retain an authorial role in a future workflow.
The disagreement is larger than a single character. Consent is one question: did a person knowingly license their likeness? Compensation is another: how are contributors paid when their face, voice or performance can be reused? Creative labor is a third: does the process create meaningful new roles, or compress jobs traditionally done by people into a smaller number of technical positions? None of those issues is solved merely by calling a character a tool.
Fountain 0 has pursued the proof-of-concept route. Its docudrama Dreams of Violets premiered at Tribeca in June and became one of the first fully AI-generated films to play at a major festival. The company also made Odysseus: The Fall, an imitation-minded exercise intended to demonstrate how close AI might come to the scale of a costly summer spectacle. The response was not universally favorable, a useful reminder that technical demonstration and artistic equivalence are separate claims.
Fountain 0 has also licensed Sunny Hostin’s likeness for AI adaptations of her romance novels. That makes licensing a crucial counterpoint to the synthetic-performer debate. A licensed likeness is not automatically free of ethical questions, but it is materially different from a system using a recognizable identity without a clear agreement.
Runway, Higgsfield and Black Forest Labs are trying to make AI part of the filmmaking toolset
Runway and Higgsfield are concentrating on platforms, contests, exhibitions and creator access. Runway has held an annual AI Film Festival since 2022, while Higgsfield has organized competitions and likeness-licensing arrangements. Their bet is that new formats emerge when more people can turn a script or intellectual property into video without the traditional scale of a feature production.
This is the democratization argument: lower barriers can give artists with fewer resources a chance to attempt stories previously out of reach. It is a real promise, but not a guarantee of better storytelling. A tool that makes images easier to generate cannot by itself provide taste, emotional clarity, production discipline or a coherent point of view.
Black Forest Labs represents another form of validation. Martin Scorsese’s involvement with the German company has been connected to research and exploration, including help plotting a storyboard for an upcoming film. The company has not said whether its Flux models are being used on his next feature. That uncertainty should remain intact; a high-profile partnership is evidence of exploration, not confirmation that an AI tool is embedded in a finished Scorsese production.
For viewers who are wary of digitally altered visuals in games as well as film, the most useful standard is to judge the actual result while asking clear questions about process and consent. That same pause-before-verdict instinct is relevant when considering unfinished or partial material, as in this look at why early datamined game models should not be treated as final.
Public trust may be Hollywood’s real bottleneck
Social-media criticism of AI use in films including Eli Roth’s Ice Cream Man and David Dastmalchian-led Late Night With the Devil shows that audience resistance is not abstract. Viewers are concerned about perceived “AI slop,” the dismissive term for output seen as cheap, careless or devoid of human creative intent. They are also concerned about the wider social effects of AI.
Jon Erwin, whose Innovative Dreams venture was launched with AI firm Luma, argues that productive uses are too often bundled together with low-quality output. His projects have incorporated generative AI, including Amazon MGM’s House of David and Angel Studios’ Young Washington. His position reflects an important pro-AI case: filmmaking has long combined art with disruptive technology, and artists can use new tools in surprising ways.
But the skeptical response is equally practical. If the industry wants audiences to distinguish carefully directed, consent-based work from low-effort material, it will need to make that distinction visible. Human oversight, licensed identities, rights-respecting training practices, honest communication and creative accountability are not peripheral public-relations details. They are the conditions under which experimentation has a chance to earn legitimacy.
Hollywood’s AI competition is therefore not simply a contest to make content faster or cheaper. It is a contest to establish a model of authorship that creators, performers and viewers can live with. The studios and startups moving first may gain technical experience, but the winners will be the ones that can demonstrate that more capable tools do not require less human control.
