OpenAI says it’s killing Sora.
We’re saying goodbye to Sora. To everyone who created with Sora, shared it, and built the community around it: thank you. What you did with Sora matters and we know this news is disappointing.
We’ll share more information soon, including timelines and details for the app and API…
– Sora (@soraofficialapp) 24 March 2026
I wrote a week ago that Sora was possibly at a tipping point amid OpenAI’s pivot to business and productivity tools. Looks like he was serious. It will be gone soon. Despite Sora’s ability to generate headlines (when it was first previewed, we at Gizmodo called it “breathtaking, yet terrifying”) the company is holding off on this compute-guzzling AI video experiment.
For my part, on that first Sora day in February of 2024, I was driving on Interstate 10 in the Mojave Desert when I first saw Sam Altman’s tweets about OpenAI’s unreleased video models. The technology he was demonstrating felt like such a huge and sudden leap in capability that I had to pull over and stare at my phone.
That was my peak moment of AI vertigo. I’ve never felt such a powerful visceral reaction to a piece of AI technology, and it’s doubtful I ever will again. Partly because my brain took some getting used to, and figuring out the slope became a new survival skill. It didn’t hurt that some of the early outputs published by OpenAI were bizarre, unexpected failures.
Altman revealed that the model was to be called Sora from the beginning, but then OpenAI let the brand lie dormant for several months. Other AI video generators were fully released to the public, and then in September last year, OpenAI confusingly released Sora 2. But it also provided the Sora brand name to OpenAI’s new TikTok-like video sharing app, which became OpenAI’s consumer-facing access point to that once-jaw-dropping video model. The most important feature in the Sora app was essentially the option to deepfake yourself and allow others to deepfake you.
The results were so horrific that I couldn’t look away.
Against our better judgment, several reputable commentators – and yours truly – were briefly trapped in Sormania. Letting models have their way with your image was a bit like allowing children on a sugar high to color you with markers and glitter, minus the sensation of human connection, and with a much more tangible reputational risk.
But the thrill wore off, and the social aspects of the app never became a daily habit for that first wave of users. It was rumored for some time that OpenAI was going to convert Sora into ChatGPT, but this never happened. Now Sora is on death row, waiting to be killed.
As of press time, it was still possible to watch and create videos with the Sora app. The official Sora Disney has already backed out of its content-sharing agreement with OpenAI.
Gizmodo contacted OpenAI for clarity on what this means for the continued existence of the model. Although shutting down the video-sharing app is straightforward, it’s less clear whether the core model will be folded into another model, protected in some other way, or wiped off the face of the earth. We’ll update if OpenAI gets back to us.
<a href