OpenAI’s Sora video tool was leaked by a group of artists to protest ‘corporate AI overlords’
A group of artists volunteering as beta testers for OpenAI’s new and unreleased AI video product, Sora, publicly shared access to the tool on Tuesday to protest what they said were the company’s exploitative practices.
In an open letter posted on the Hugging Face platform and addressed to “corporate AI overlords,” the group published the access to Sora’s API, the special code that allows outside users to access the tool.
The group claimed they were invited by OpenAI to serve as Sora “testers” and “creative partners,” but realized they were being “lured into ‘art washing’ to tell the world that Sora is a useful tool for artists.” They also maintained that OpenAI controlled and approved all outputs, and that the early-access program appeared to be “less about creative expression and critique, and more about PR and advertisement.”
Sora is one of the most anticipated new products from OpenAI, the privately held company that makes ChatGPT and which has been valued at $157 billion. Generative AI video technology has the potential to upend Hollywood and many other creative industries that rely on video creation, from advertising to art. As word spread online about access to the Sora model on Tuesday, users quickly began posting their video experiments. “It can be shut down anytime, try it now!” said one excited user on X. “It can generate 1080P and up to 10s video! And the results are incredible!”
Following the leak, an OpenAI spokesperson confirmed that the company has “temporarily paused user access while we look into this.” The spokesperson emphasized that Sora is still in a research preview, and that “hundreds of artists” have shaped Sora’s development, helping prioritize new features and safeguards. “Participation is voluntary, with no obligation to provide feedback or use the tool,” the spokesperson said. “We’ve been excited to offer these artists free access and will continue supporting them through grants, events, and other programs. We believe AI can be a powerful creative tool and are committed to making Sora both useful and safe.”
Marc Rotenberg, founder and executive director of the Washington, D.C.–based Center for AI and Digital Policy, pointed out that the leak was “deeply ironic” since OpenAI was first established as a company whose research was open to all. “It’s the reason that Elon Musk put money into it, and it was the subsequent commercialization that explains the reason he became disenchanted,” he said. “So if you go back to his mission, I think you would celebrate what the artists did in this moment. But of course, if you’re Microsoft, and you just poured in $10 billion to further your proprietary model, this is probably not a good day,” Rotenberg said, referring to Microsoft’s partnership with OpenAI.
It’s important to note that the news of the API access is much less devastating than if the entire Sora model—including the code and weights (which serve as the “brains” behind the model) had been leaked. OpenAI quickly shut down access to the leaked API on Tuesday.
Still, the fact that the company’s own beta testers are protesting is notable: Over the past several years, artists have voiced growing concerns about exploitative practices in the realm of generative AI, particularly around issues of copyright and content usage. Generative AI models often rely on vast datasets scraped from publicly available digital content, much of which includes artwork, illustrations, and other creative works created by artists.
In addition, AI-watchers have been extremely curious to find out how the highly anticipated OpenAI Sora model performs. In February, OpenAI released several high-definition video clips generated by Sora, but there have only been occasional demo videos released since then.
