Last September, when OpenAI quietly released the Sora 2 app to the public, the discourse around it was not quiet at all. Commentators who had spent months watching the model’s preview clips, the golden retriever bounding through autumn leaves, the Tokyo street scene that looked almost real, lined up to declare that something had shifted.
Not incrementally. Fundamentally. A technology that could conjure moving images from a sentence of text was, the argument went, the beginning of the end for a certain kind of human work.
The actors, the animators, the cinematographers: their skills, accumulated over careers, were suddenly described as provisional. Hollywood unions put out statements. Illustrators circulated open letters. The word “replacement” appeared in so many headlines that it began to feel less like a prediction and more like a weather forecast, delivered with the same casual authority.
On March 24, 2026, OpenAI announced it was shutting Sora down. The app had been live for six months. It had not replaced anyone.
This matters not because Sora was a failure in any technical sense. The underlying model was, by most accounts, genuinely impressive: capable of generating video that could fool a casual viewer, paired with audio generation that made the illusion more convincing still.
The 💜 of EU tech
The latest rumblings from the EU tech scene, a story from our wise ol’ founder Boris, and some questionable AI art. It’s free, every week, in your inbox. Sign up now!
OpenAI’s own communications described Sora 2 as potentially “the GPT-3.5 moment for video,” a comparison meant to signal a threshold crossed. It reached a million downloads faster than ChatGPT had. For a few weeks, it sat at the top of Apple’s App Store.
What it could not do was make people want to come back.
By January 2026, downloads had fallen 45%. The social layer that OpenAI had built around it, a TikTok-style feed of AI-generated clips, never congealed into a daily habit.
The company told employees this week that shutting down Sora would free up compute resources for its next generation of models. The research team, it said, would pivot to “world simulation research” in service of robotics. The pivot, in other words, was away from creativity and toward utility.
That pivot is worth sitting with, because it runs directly against the grain of everything the AI-will-replace-artists narrative assumed. That narrative rested on a specific theory of creative work: that the value audiences find in art, film, and performance lies primarily in the output, the image, the sound, the story, and that if a machine can produce a convincing version of that output, the human who used to produce it becomes redundant.
Sora was supposed to be the proof of concept.
What Sora revealed instead was the other half of the equation that the replacement theorists tended to ignore. Audiences do not simply consume outputs. They engage with them in relation to their origins.
The deepfake of a dead civil rights leader performing a comedy sketch is not the same cultural object as a film about that leader’s life, even if the pixel-by-pixel quality of the former is technically superior.
When families of Martin Luther King Jr. and Robin Williams went public to object to AI videos made in their likeness on the Sora platform, they were not making an abstract ethical argument. They were articulating something that users felt too, in the moment they swiped past yet another AI clip: that the context of creation is part of what makes a creative work worth caring about.
OpenAI understood this dynamic, or at least sensed it, which is why it pursued the Disney deal with such urgency.
A licensing agreement covering more than 200 Disney, Marvel, Pixar, and Star Wars characters was not just a commercial arrangement. It was an attempt to borrow cultural legitimacy, to make AI-generated video feel attached to something that audiences already had a relationship with.
The deal was announced in December 2025. It was unwound this week, with Disney noting that no money had changed hands.
The pattern here is legible. OpenAI tried to build an audience for AI creativity by making the output impressive, then by making it personalisable (the “cameo” feature, in which users could upload their own likenesses and let strangers generate deepfakes of them, was the app’s flagship attraction), then by grafting it onto existing intellectual property with established fanbases.
None of it produced sustained engagement. The app that was going to replace creative professionals could not hold the attention of the people it was meant to entertain.
None of this means that AI video generation is without consequence for the creative industries. The tools exist; they will be used; some of that use will displace work that human beings used to do.
These are real concerns, and the unions and guilds raising them are not wrong to push for contractual protections. But the specific claim that dominated coverage of Sora’s launch, that AI was on the verge of making human creativity redundant, that actors and artists were facing structural obsolescence, has run into the oldest problem in the history of cultural technology: people do not simply want content.
They want content made by someone.
The streaming era was supposed to kill cinema. It made theatrical releases more selective, but it did not kill the desire to see a film on a large screen, with other people, made by a director whose previous work you knew.
The MP3 was supposed to kill music. It restructured the industry catastrophically, and yet live performance has grown.
Each wave of reproductive technology reshapes the economics of creative work without eliminating the appetite for the human hand behind it.
Sora’s six-month lifespan does not prove that AI cannot be a useful tool for filmmakers, animators, or storytellers. It proves something narrower and perhaps more important: that AI-generated creativity, presented on its own terms, as the point rather than the instrument, did not find its audience.
The people who were supposed to be replaced turned out to be the ones whose absence the audience noticed.
OpenAI’s next chapter, by its own account, is in robotics and enterprise productivity. It is a telling retreat, toward machines that assist human work rather than perform human expression.
Perhaps that is where the genuine utility lies. The artists, for now, are still here.
Stephan is the sports journalist for the Maple Grove Report.
