OpenAI's still unreleased flagship AI video generator Sora was leaked to HuggingFace yesterday. Disgruntled artists who were given early access decided to share it with the world in protest against its use as a “PR puppet.”

Sora was first unveiled to the world in February of this year; Kling, Runway, Luma Labs, and Hailuo have since begun to catch up, but at the time it was well ahead of any other AI video generator.

In response to leaks and criticism, OpenAI quickly shut down access to all users, stating that no one with early access was obligated to use it or provide feedback.

Instead of releasing Sora to the public as is, OpenAI went through a rigorous red team and research preview process. This included releasing it to a select group of industry experts and creators. The goal was to improve the overall user experience while ensuring that no misinformation, illegal content, or videos of real people could be created.

Outraged by this slow rollout and process, a group of artists decided to protest by sharing their access keys on HuggingFace for everyone to try. They wrote 'We received access to Sora on the promise that we would be early testers, red-teamers, and creative partners.' The group says they felt invited to “wash the art” and forced to tell the world that Sora is “a useful tool for artists.” [Access to the leaked Sora was immediately removed. In fact, OpenAI was forced to remove access to anyone who had permission to try Sora. In a statement sent to Tom's Guide, the company said that participation in the alpha program was voluntary and that there was no obligation to use the tool or provide feedback.

The spokesperson further explained that “the hundreds of artists who participated in our alpha helped shape Sora's development and prioritize new features and safeguards,” and that “we are excited to be able to offer free access to these artists and look forward to continuing to provide them with the tools and feedback they need to make Sora better, continue to support them through grants, events, and other programs. We believe AI can be a powerful creative tool, and we are committed to making Sora convenient and safe."

Sora is not ready for a primetime release; OpenAI describes it as a “research preview,” but by all accounts it is still ahead of its competitors in terms of movement, visual clarity, and overall quality. It also appears that a new Turbo mode has been added, which may have fixed some of the problems related to slow generation.

“Sora is still in the research preview stage and we are working on balancing creativity with robust safeguards for wider use,” an OpenAI spokesperson told me.

As part of Sora's research preview period, OpenAI is also providing access and in some cases funding for projects like the Tribeca Film Festival's Sora Shorts program and exhibitions at Heron Arts and the New York Gallery.

The only responsibility that artists with early access to Sora had to uphold was not to share confidential information while it was still in development. They were also encouraged to provide candid feedback.

I am eagerly awaiting Sora's release and feel that OpenAI is being too cautious in its slow rollout, given the increased competition. However, in my view, if we are given early access, the time to share criticisms of the actual platform is when it goes live. That way, others can tell the full story once they have access.