Everything you need to know about OpenAI’s war against Sora whistleblowers

OpenAI is going through a controversial and extremely sensitive situation with a “voluntary” group of testers for its generative video AI tool Sora. All this, after several months of this testing dynamic, a collective calling itself PR Puppets said that they were asked to work for free as bug testers and also do research and development, while simultaneously participating in the “artistic laundering” of artificial intelligence tools.

But OpenAI says participation in early alpha testing is voluntary and does not require testers to provide feedback.

According to Git commit logs, PR Puppets published their “Generate with Sora” hotspot on Hugging Face around 8:30 ET on Tuesday morning. Artificial intelligence experts on social media quickly noticed this post and confirmed that the page is connected to OpenAI’s real Sora API endpoints and hosted on the videos.openai.com domain, presumably with authentication tokens provided to testers by OpenAI itself.

Of course, after a few hours this was all quickly neutralized, but it only took a few hours for many users to start leaving their creations online:

Oh my god, OpenAI Sora leaked online!

Free to use now on Huggingface, link in comments

It can be closed at any time, try it now! It can generate 1080P videos up to 10 seconds long! And the results are incredible!

9 examples: pic.twitter.com/rIJJv5TQTo

— el.cine (@EHuanglu) November 26, 2024

An OpenAI spokesperson told The Washington Post that the company is temporarily suspending all test access to Sora to assess the situation. Other users have dug into the code to find clues about the different modes and “styles” that might be in development for Sora.

The truth about the Sora testers’ lawsuit

In the letter, the PR Puppets team emphasized that their grievance was the deception associated with unpaid participation as covert work to improve the application.

In an open letter to their “corporate AI overlords,” the PR group Puppets said it was protesting on behalf of the roughly 300 artists who were given early access to Sora only to “provide unpaid labor through bug testing, feedback, and experimental work.” under the program for a company worth $150 billion. [OpenAI]” The group said few of these artists would see their work realized in the wider release of Sora-produced films, and complained that OpenAI required approval before any Sora alpha output could be released publicly.

“We are not against the use of artificial intelligence technology as an artistic tool (if we were, we probably would not have been invited to this show),” PR Puppets wrote. “What we disagree with is the way this artist program has been implemented and how this tool is being framed ahead of a possible public launch. “We’re sharing this with the world in hopes that OpenAI will become more open, more artist-friendly, and support the arts beyond PR stunts.”

An OpenAI representative spoke with ArsTechnica and responded to an email sent by people from the team.

“Sora is still in the research phase and we are working to balance creativity and strict security measures for wider use. Hundreds of artists in our alpha influenced the development of Sora, helping to prioritize new features and security measures. Participation is voluntary, with no obligation to provide feedback or use the tool.

Regarding the open and official availability of Sora to users, OpenAI CPO Kevin Weil said at a recent Reddit AMA that the platform’s implementation has been delayed due to “the need to refine the model, the need to get good at security, spoofing and other things.” and the need to scale computing.”