OpenAI suspends access to Sora video generation tool after artists protest
Earlier this year OpenAI unveiled Sora, a text-to-video AI model, showing off detailed scenes and complex camera motion from relatively simple prompts. It's been radio silence since then, but the company recently granted artists free early access to the tool for testing. However, a group of around 20 of those just leaked access to Sora in protest, saying they were acting as "PR puppets," prompting OpenAI to suspend access, The Washington Post reported.
"We received access to Sora with the promise to be early testers, red teamers and creative partners. However, we believe instead we are being lured into 'art washing' to tell the world that Sora is a useful tool for artists," the group wrote on the AI art repository site, Hugging Face.
Pointing out that OpenAI recently hit a $150 billion evaluation, they noted that artists have been providing unpaid testing and feedback. The group also noted that all Sora-generated content needed to be approved by OpenAI, making it "less about creative expression... and more about PR and advertisement." The group then said it had released the tool to let anyone play with it, saying it hopes that OpenAI will "support the arts beyond PR stunts."
In response, OpenAI shut down early Sora access after just three hours while it looks into the situation. "Hundreds of artists in our alpha have shaped Sora’s development, helping prioritize new features and safeguards,” OpenAI spokesperson Niko Felix wrote in a message to The Post. “Participation is voluntary, with no obligation to provide feedback or use the tool.” Another artist in the program, André Allen Anjos, chimed in as well saying that the protest artists' stance didn't reflect the views of most artists in the program.
Though Sora isn't yet widely available, the tool has been scrutinized over its training materials. In March, OpenAI CTO Mira Murati said she wasn't sure if Sora took training data from YouTube or other video platforms. The next month, YouTube's CEO specifically warned OpenAI that training models on its videos was against its terms of service.
