OpenAI has launched its AI video generator Sora with restrictions on depicting people to prevent misuse such as deepfakes.
- OpenAI has released Sora, a new AI video generator.
- Users with a premium ChatGPT version can create videos based on text prompts.
- Example videos include fictional scenarios, like bears wrestling and cats drinking coffee.
- Access to human depiction is limited to a select group of invited users due to concerns about misuse, such as misappropriation of likeness and deepfakes.
- OpenAI emphasizes its commitment to preventing harmful uses, including child sexual abuse material and sexual deepfakes.
- The service has temporarily paused new account creation due to high demand upon release.
- OpenAI’s focus includes engaging with artists and policymakers prior to public release.
- Concerns have been raised regarding the potential for AI video tools to impersonate real people, particularly in political contexts.
- OpenAI has set strict content filters to block nudity and other inappropriate materials.
- The company has not disclosed the sources of imagery and video used for training the Sora model.