Responsible Launch of Sora

Sora 2 and the Sora app deliver state-of-the-art video generation technology while integrating robust safety measures from the design stage.

The key safeguards are as follows:

• AI Content Identification: Every generated video includes a visible watermark and C2PA metadata. In addition, internal reverse image and audio search tools allow accurate tracing of whether content was produced by Sora.
• Likeness and Character Protection: The Sora characters feature lets users directly control their own likeness. Depictions of public figures are blocked, and users can review or delete videos that include their character.
• Teen Protection: For minor users, exposure to adult content is restricted, and ChatGPT's new parental controls allow management of message reception and feed settings.
• Harmful Content Filtering: Prompts and outputs are inspected from multiple angles to block sexual content, terrorist propaganda, self-harm promotion, and more. Red teaming is used to prepare for emerging risks, combining automated systems with human review.
• Audio and User Rights: Imitation of real artists' music is blocked, and users can decide for themselves whether to share generated content, as well as block or report inappropriate accounts.