OpenAI’s Sora Video Tool Leaked Amid Accusations of Exploitation

OpenAI’s unreleased text-to-video tool, Sora, was leaked through a Hugging Face frontend by a group protesting its development practices. They accuse OpenAI of exploiting testers, alleging that hundreds of artists have provided unpaid feedback under the guise of creative collaboration, with OpenAI using this to bolster its public image.

The leaked tool, which allowed users to generate short, high-resolution videos from text prompts, was quickly disabled after the breach. However, sample videos featuring realistic animations like a walking woman and a neon-lit cityscape sparked widespread discussion online.

The group, calling themselves “Sora PR Puppets,” claims that OpenAI unfairly pressures early testers to create a positive narrative while maintaining tight control over what’s shared publicly. They argue that this approach prioritizes marketing over genuine artist collaboration and call for more ethical development practices.

This controversy revives the discussion about whether AI should be trained on artists’ work without explicit consent.

OpenAI denies the accusations, stating that participation in Sora’s testing is voluntary and that the company is focused on balancing innovation with safety. However, the incident has reignited debates over fair compensation for artists whose work informs AI training and whether proprietary AI models can be ethical tools for creative professionals.

X user “Chubby” shared some videos which were reportedly generated with this leaked model before it was taken down. Those were kind of underwhelming, compared to the current best examples on the market.

Read in details: https://techcrunch.com/2024/11/26/artists-appears-to-have-leaked-access-to-openais-sora/