It started as a joke. Not literally, but close enough. The early attempts at AI-generated imagery in film and television arrived with all the grace of a drunk uncle at a wedding, lightly unnerving, a bit creepy, and very easy to dismiss. The uncanny valley was more of an uncanny Grand Canyon.
People watched, laughed a bit nervously, and agreed that while it was technically impressive in a PowerPoint-presentation kind of way, it was never going to threaten the real thing, right?
If you need a single exhibit for the prosecution, the 2023 viral video of Will Smith eating spaghetti. Generated by an early AI video model and shared widely online, it depicted the actor in a vaguely recognisable form enthusiastically consuming pasta in a manner that no human being has ever consumed pasta, or anything else, in the history of eating. The noodles behaved like sentient creatures, the fork seemed to be improvising being a fork. Nightmarish as the video was, it was also, in retrospect, a significant cultural moment, the point at which a lot of people first clocked that something genuinely strange was happening with this technology.
That consensus lasted about eighteen months.
What happened next was the kind of technological acceleration that tends to blindside entire industries. The tools got better at a pace that defied all reasonable expectations, and what had looked like a novelty just a couple of years earlier was, by 2024, producing results that gave working professionals genuine cause for concern, not because AI had become indistinguishable from human craft, but because it had become good enough. Good enough to fool a casual eye. Good enough to undercut a budget. Good enough to make studio executives start asking uncomfortable questions in rooms that writers, directors, and visual artists were no longer invited to attend. The industry pants-shitting, metaphorical and, in certain industry circles, allegedly otherwise, began.
The Trailer Problem
Nowhere has AI’s disruptive potential been felt more visibly than in the world of filmmaking and marketing. The most visible and chaotic example is the rise of AI-generated fan trailers. These increasingly convincing mock-ups of unannounced sequels, reboots, and genuine forthcoming films have created a crisis of audience trust that the industry is only beginning to reckon with.
When a fake AI trailer for a hypothetical film drops on YouTube and racks up three million views over a weekend, the imagery sticks. I got into a spirited discussion with several people on TikTok who believed the AI slop trailer they saw for 28 Years Later, barely a month after the project was announced, so not a frame of footage had been shot, was real. Despite pointing out that production hadn’t started and that it was clearly a fake, they simply wouldn’t accept anything less than what they had seen was 100% legit.
This problem doesn’t even go away when the real trailer, for the real film, made by real people, eventually lands, as a significant portion of the audience approaches it with scepticism baked in. Given the overall condition of the movie industry presently, that’s not something studios should take lightly. Comment sections fill with people confidently declaring the official material AI-generated when it isn’t. Studios find themselves in the absurd position of having to prove authenticity for work that is entirely human-made. This is not a minor inconvenience. It is a fundamental erosion of the relationship between a film and its potential audience, the moment of first contact that marketing teams spend months engineering, undermined before it has a chance to land. I’ve already lost count of the AI trailers for Avengers: Doomsday.
The Disney-Sora Situation
For those unfamiliar with what Sora 2 actually represents: it is, in simplified terms, a system that can produce high-quality video footage from text prompts. You describe a scene, the lighting, mood, movement, composition, and the model generates it. No cameras. No locations. No crew. The implications for production costs are significant enough that you do not need to be a film industry insider to understand why studios are interested and why everyone else is alarmed.
Which brings us to the deal that seemed, for a brief period, to signal where everything was heading. Earlier this year, news emerged that Disney had entered into a significant agreement with OpenAI centred around Sora 2, the video-generation model that had already demonstrated a capacity for cinematic imagery that stopped a lot of people cold. The details, as they filtered out, suggested a level of institutional commitment that felt like a line being crossed, one of the most powerful studios in the world formally embracing generative AI as a production tool, not merely a curiosity.
Then, with a swiftness that made even cynical industry observers blink, it collapsed. As recently as the Monday before the announcement, teams from Disney and OpenAI had been in a meeting about the Sora project. Thirty minutes after that meeting ended, the Disney side was informed that OpenAI was discontinuing the video app entirely. Disney’s planned $1 billion investment in OpenAI went with it. According to Reuters, that was that. Some would have us believe that the Sora 2 episode is a dead end, that it’s all fine now because this deal didn’t go through. This was more like a warning shot: this is happening, the industry knows it, and nobody has agreed on the terms yet. Some corners of the industry want to work with AI; the rest want to kill it with fire, bury it, and salt the ground.
The Affleck Angle
Not every deal collapsed. In March 2026, Netflix acquired InterPositive, the AI filmmaking tool company founded by Ben Affleck, for a reported $600 million. Where Sora was about generating footage from nothing, InterPositive takes a different approach: it builds proprietary AI models trained on footage a production has already shot, using that data to solve the logistical and technical problems that slow filmmaking down. Affleck described it as taking out all the difficult, technical stuff that gets in the way of actually making the film.
Netflix’s chief content officer Bela Bajaria was careful to frame the acquisition in the language the industry needed to hear, insisting that new tools should expand creative freedom, not constrain it or replace the work of writers, directors, actors and crews. In the same week, Affleck’s production company Artists Equity, co-founded with Matt Damon, signed a separate multi-year deal with Netflix to develop and distribute future projects. Whether the reassurances hold once InterPositive is fully embedded in Netflix’s production pipeline remains to be seen. But the combination of a $600 million acquisition and a production partnership signed in the same breath left certain corners of the industry feeling less like they were watching innovation and more like they were watching a consolidation.
What Can Actually Be Done
The honest answer is: not much. Unions are still fighting over the wider implications – the WGA and SAG-AFTRA strikes of 2023 produced landmark agreements that included, for the first time, explicit AI protections for writers and performers. Those agreements matter. They establish a much needed precedent, but even that has a limited shelf life when the technology moves faster than contract cycles.
Watermarking and provenance tools, systems that would allow audiences and platforms to verify whether content was AI-generated are in development and represent perhaps the most realistic near-term intervention. The EU AI Act is already pushing mandatory disclosure requirements. Whether that translates meaningfully into the chaotic ecosystem of social media, where most of the damage is done, remains to be seen.
The Actual Reckoning
Let’s be honest about both sides of this, because the discourse tends to flatten nuance in both directions. The legitimate uses are real. AI tools are already helping smaller productions achieve visual effects that would previously have been financially impossible. De-ageing technology, when used with consent and craft, has produced genuinely remarkable results. That said, it would be great if digital recreations of dead actors could stop being a thing. Translation and dubbing workflows are being transformed in ways that could meaningfully expand the global reach of non-English language cinema. These are not trivial benefits.
But the concerns are also real, the most immediate is labour. Visual effects artists, a workforce that has historically been undervalued and overworked even in the best of times, are facing displacement at a scale that existing safety nets were not built to handle. Voice actors are finding their performances cloned without consent. Background performers are being replaced by generated crowds. Together, they represent a structural shift in who gets paid to make films and television, you won’t be surprised to discover the people at the top rarely take a pay cut, it’s always those that actually make the movie or TV show that ultimately suffer.
The World’s First (but not last) AI Actor
Then there is Tilly Norwood, a world first and a constant point of friction in Hollywood. Created by Xicoia, the AI division of London-based production company Particle6, founded by Dutch actress-turned-producer Eline van der Velden, Norwood is billed as the world’s first AI actor. She has an Instagram account, a growing fanbase, and a creator who has stated, with apparent sincerity, that she intends Norwood to be the next Scarlett Johansson. Particle6 claims that using Norwood can cut production costs by 90%. Hollywood agents were reportedly in talks to sign her. SAG-AFTRA responded with the kind of statement that suggests an organisation that is genuinely alarmed, describing Norwood not as an actor but as “a character generated by a computer program that was trained on the work of countless professional performers, without permission or compensation.” Van der Velden has since announced the “Tillyverse”, a rapidly expanding digital universe in which Norwood and a new generation of AI characters will “live, collaborate and build careers.”
The subtler concern is creative. When the tools that generate imagery are trained on existing human work, often without the knowledge or compensation of the people whose work was used – what gets produced is, at its root, a sophisticated recombination of everything that already exists. It can be technically impressive, on that we can all agree. However, as impressive as it may be, it cannot, by definition, surprise you in the way that genuine creative vision can. The risk is not that AI replaces human storytelling. It is that it gradually, almost imperceptibly, crowds it out.
Where This Ends Up
Nobody knows, and anyone who tells you otherwise is selling something, possibly generated by a model trained on sales copy. I am as disappointed as you are that we have reached this stage in the article and have no real answers, but that’s kind of the point.
Here is a useful way to measure how far we have already come, and just how concerned we should be in all corners of the creative industry. Remember the Will Smith spaghetti video, the 2023 nightmare of sentient noodles and an approximation of a face that had clearly never met pasta before? There is a more recent version, and this time it is, for the most part, ‘convincing’. That is a demonstration of how quickly the gap between “obviously fake” and “plausibly real” can close when the people building these tools are motivated and well-funded.
The two videos, taken together, tell you almost everything you need to know about where this is heading and why the stakes are what they are. In two years, AI video went from a viral laugh to something that requires a second look. That recent video of an AI Brad Pitt and Tom Cruise fighting on a rooftop was a clear indication that the technology is rapidly evolving. Where will AI be in 2028? Multiple new AI studios are making completely AI generated animated films, so what does seem certain is that the question is no longer whether AI will find a place in Hollywood, but what kind of place, on whose terms, and at what cost. The studios want efficiency. The creatives want the protections that give their industry stability. The audiences, increasingly, just want to know what they are actually watching, and are losing the ability to tell.
AI is not going back in the box, the spaghetti was always going to get better. The only question left is who decides what gets made with it, and whether the answer to that question will be one the film industry, or any of us, can live with.
