It’s been just under two weeks since OpenAI launched ‘Sora 2’, its newest AI video generation tool on September 30.
But more surprising and unusual was the launch of the accompanying short-form video social media site. The app has already gained more than a million users. This is despite it only being available in North America exclusively on the Apple app store, as well as being invite-only.
The site clearly has appeal but it perhaps seems out of character for a company that has been entirely dedicated to AI research and sees itself as bringing advanced AI to the world to now be throwing its hat in the ring of social media.
There’s also been great concerns surrounding ‘AI Slop’, the nickname given to short-form and incredibly low-effort AI videos that are beginning to flood all social media sites, as well as concerns around potential misuse of the highly-realistic looking technology.
So today I ask the questions on everyone’s mind: Why make this and who is it for?
The Financial Motive
OpenAI was not the first to dream up AI short-form content. Mark Zuckerberg’s Meta had launched a similar site called “Vibes” just days prior and social media companies like Youtube and Tiktok had already begun to incorporate AI tools into their platforms.
The business appeal of AI social media is understandable. It could allow anyone, not just specialised content creators, to create videos that users wanted to watch. AI could theoretically tailor content for anyone’s taste. Using generative AI also promises an ‘infinite’ stream of content which is bound to keep people on these apps for longer periods of time
However, all of the other companies that are trying to make AI short-form video work are already in the social media business, so why would OpenAI try its hand at making it?
The launch of Sora seems to be just a small part of recent efforts by OpenAI to increase their offerings and revenues. They also launched ChatGPT integration with Spotify and other apps on October 6.
These new launches may have something to do with the recent deals OpenAI have made with other tech companies such as Nvidia and AMD. Those deals are now worth more than $1 trillion which are massive financial obligations for a company that has only ever managed to lose money making AI products. OpenAI continues to burn through cash, reporting a $7.8 billion loss in the first half of 2025 alone.
And so, they appear to be desperate enough to take the risk of creating a social media site on their own rather than partnering with their competitors. They’ve also begun utilising advertising on the service in a first for the company, which ordinarily relies on profits from subscriptions for its premium models.
The risk may pay off but their lack of experience in the sector has certainly become evident in the somewhat tumultuous launch of the service that has seen copyright issues and deepfake concerns with still no profits in sight.
A Clumsy Launch
OpenAI have never really been known for their respect of intellectual property. But Sora, in violation of just about every copyright law on the planet, initially used an “Opt-out” copyright system. This meant that rightsholders had to explicitly refuse to have their works on the platform rather than OpenAI not permitting copyrighted works by default as all other social media sites do.
This was soon walked back to the more traditional Opt-in system around a week later. OpenAI Ceo Sam Altman said in his blog that they wanted to “Let rightsholders decide how to proceed (our aim of course is to make it so compelling that many people want to).”
How Sam Altman plans on convincing rightsholders to accept this is unclear. Traditional media is already largely under threat from more conventional social media but AI video could unleash a torrent of brand-damaging content. The Motion Pictures Association have already condemned the site, even in spite of the measures taken to appease rightsholders.
The crackdown on copyrighted material has also frustrated users, especially those who experienced the intitial period where just about everything was fair-game. The app’s rating on the Apple app store has now dropped to 2.9 stars with many reviewers complaining about the strict content policies, especially when compared with the first week of launch.
Another issue OpenAI appears to be facing is a lack of revenue. Sam Altman ordinarily enjoys bragging about the scale of his company’s losses, but in an update posted to his blog he suggested that they needed to raise revenues. He somewhat contradicted himself in that same statement, however, by suggesting that they then split these revenues with rightsholders whose content was being used on the platform.
It’s entirely unclear how Sam Altman will be able to raise revenues to a level where it would even be possible to share them with anyone else. In the update, he says that “People are generating much more than we expected per user, and a lot of videos are being generated for very small audiences.”
This appears to suggest that users are using Sora to generate lots of AI videos that they are then posting elsewhere on sites such as Twitter or Instagram. Effectively, OpenAI is spending millions to make short-form content for their competitors and users are draining all of OpenAI’s compute in exchange for nothing.
These business failures are not just comical, they also represent a small glimmer of hope in an era of rampant AI advance. They suggest that the theoretical nightmare of unlimited AI content being generated about anything and everything and flooding the internet may not come true. They also suggest that one of AI’s biggest players is beginning to get desperate for cash.
And so the AI Bubble may not end with a crash but instead with a collage of incredibly boring 10 second AI Shorts and with the desperate cry of Sam Altman trying to sell you Ad-space next to them. That doesn’t mean that the future is so bright, however.
A horrifying glimpse into the future
Concerns around deepfakes have been around ever since AI video generation was in its infancy. But we’ve come a long way since that stage.
Sora 2 produces photorealistic content with very little of the visual glitches common to older models. This feat is impressive but it’s also terrifying to imagine what these tools could be used for if millions of people had access to them. Luckily (Or perhaps unluckily) for us, we don’t need to imagine.
As mentioned earlier, access to the app is restricted and content is heavily moderated but that hasn’t stopped some users from generating some pretty concerning material. Such as this video that depicts Sam Altman stealing GPUs from a store, a tongue-in-cheek reference to OpenAIs need for GPUs in its AI infrastructure.
Despite the comedic nature of the video, we should make no doubt that it attempts to (and somewhat succeeds in) portraying someone committing a crime. As funny as it may be when its someone as powerful as Sam Altman, the potential for this to be misused to frame people is obvious.
Another video depicts female Olympians competing in events based off of common household chores such as getting a beer from the fridge and cleaning kitchen sinks. While not too grossly offensive, it still shows that a determined enough user-base will be able to get around any restrictions no matter what and will no doubt further add to OpenAI’s headaches with the service
That’s to say nothing of what other players in the space may do. Elon Musk, for example, has specifically advertised Grok as an ‘Anti-woke’ AI model and has left its video and image generation largely uncensored. Others may also develop less-regulated video generators that could be used for things such as deepfakes.
Sora already uses something similar. Its ‘Cameo’ feature scans people’s faces so they can be used in AI videos. The idea is presumably to allow people to put themselves and their friends into these AI videos but encouraging users to use real people’s faces is also dangerous.
We can laugh for now when that feature is being used to depict Sam Altman as Skibidi toilet or stealing artwork from Studio Ghibli but I doubt it will be funny when ordinary people are being depicted committing crimes they didn’t and when millions have access to this powerful technology.
There’s also concerns about the watermarks, used to make it clear when videos have been AI generated. These watermarks have proved exceptionally easy to remove and so internet users may be unable to tell when videos have been faked, especially given the extreme advancements in how visually convincing the videos are.
Sam Altman, in his typical fashion, addressed many of these concerns when he first decided to launch the product. “We are aware of how addictive a service like this could become” he said in a statement on Twitter, shortly before explaining why he decided to release it anyways.
What this shows about the future of AI
We are looking, now, at a watershed moment, not too unlike when ChatGPT first launched. By comparison, however, I must say that Sora’s launch has been significantly less smooth and that gives me a lot of hope. Whether it ends with OpenAI having to retire it because it costs them too much money or with rightsholders suing them into the ground, the foundations for Sora already appear shaky.
OpenAI promises us cures to cancer and Artificial General Intelligence but it is now a company whose work is dedicated to letting you cheat on your homework and generate images of its CEO’s head in a toilet. No one knows how the future of AI will pan out but it’s safe to say the present reality gives me little fear that they’ll be taking over the world anytime soon.
These expensive projects are flashy and they are ‘fun’ but they are nowhere near the grand promises Sam Altman keeps making. And after the somewhat muted launch of GPT-5, it appears as though one of the world’s most powerful companies is beginning to run out of steam to keep the AI-hype train going.
Its latest attempt has proven to be an infinite stream of mundane and boring AI ‘Slop’ that hardly seems to prove that AI will be the future of video creation. It’s boring, it’s soulless and people simply aren’t interested.
‘AI Slop’ sounds like one of those sad things that we have to accept in society but the model clearly isn’t working. Very few are using the app to watch content and the videos that are being watched elsewhere are mostly highlights from the first week of lax content moderation.
This isn’t a case of saying “AI slop sucks, Human slop reigns supreme!” but rather a mild and cautious celebration of seeing one of the most single powerful companies in the world make a desperate misstep into short-form video creation. AI video generation is still dangerous but Sora isn’t.
So, who asked for this? No one, and no one wants it either. OpenAI is beginning to realise that.
Featured Image Credit: Sora