OpenAI's Sora model got "leaked"—non-beta testers could try it for three hours

OpenAI’s unreleased yet hotly anticipated video-generating model, Sora, has just been briefly exposed to outsiders by a group of artists who are beta testers in what could be the biggest shock in the AI sphere recently.

The artists, granted early access to Sora, accused the $157 billion-valued company of exploiting their unpaid labor for PR purposes while offering minimal compensation. In protest, they created a webpage allowing public use of Sora and posted it on HuggingFace, an open-source repository for AI models.

“While hundreds contribute for free, a select few will be chosen through a competition to have their Sora-created films screened — offering minimal compensation which pales in comparison to the substantial PR and marketing value OpenAI receives,” the protesters say.

OpenAI then quickly took it down after only three hours, and all the videos in the HuggingFace post have also been removed. The AI company muted users in its Discord channel who mentioned anything about the exposed front end.

“We are not against the use of AI technology as a tool for the arts (if we were, we probably wouldn’t have been invited to this program). What we don’t agree with is how this artist program has been rolled out and how the tool is shaping up ahead of a possible public release,” the open letter reads further.

In September, The Information learned that OpenAI had been working on an updated version of Sora that addresses technical limitations, including long generation times, style inconsistencies, and errors in physics and anatomy. That was months since its first teaser back in February.

The company trained the new version on millions of hours of diverse, high-resolution video data, but no release timeline has been confirmed.