By PYMNTS | October 12, 2026
OpenAI is reportedly considering a plan to resurrect its video model Sora.
The company is in discussions with a startup that is developing artificial intelligence (AI) video for filmmakers, and that wants to access Sora and use it to sell software and other services, The Information reported Monday (Oct. 12), citing a source familiar with the matter.
The startup was co-founded by Hollywood mogul Jeffrey Katzenberg, former Sora research lead Bill Peebles, and ex-Dropbox CFO Sujay Jaswa, the report said. The Information added that the name of this startup is unclear. Katzenberg and Jaswa’s LinkedIn profiles list them as the co-founders of the tech investment firm WndrCo.
Also unclear is whether this startup will be given access to Sora’s weights, or the settings that shape how the model responds to queries, The Information added. If the deal proceeds, the startup would still get deep enough access to substantially customize it, the source added.
The Information noted that while AI developers typically guard their model weights closely, sharing weights wouldn’t be surprising, as Peebles knows how the model works. Working with him makes sense, the report added, considering OpenAI’s initial investment in Sora.
We’d love to be your preferred source for news.
Please add us to your preferred sources list so our news, data and interviews show up in your feed. Thanks!
OpenAI introduced Sora in 2025, but began winding down its video AI efforts earlier this year as part of a larger effort to scale down its “side quests” as it focused more on core business lines and readied itself for an initial public offering tied to enterprise applications.
In other video AI news, PYMNTS spoke recently with Jay Dang, CEO and co-founder of Kaon AI, which—per Dang—avoids algorithmic recommendation entirely.
Instead of surfacing the best existing content, the platform generates new video in real time, crafting a story world designed around a user’s interests as they interact with it.
“If the perfect piece of content for the user was never made, the algorithm cannot help the user,” Dang said.
The distinction, PYMNTS wrote, is the whole business behind Kaon. Recommendation systems can only point to the closest available match, and when nothing in the library lines up, the system is out of options.
“Kaon removes that ceiling by making the content on the spot, based on what the user chooses in the moment,” the report added.
For all PYMNTS AI coverage, subscribe to the daily AI Newsletter.