OpenAI has launched Sora 2 and the Sora app with safety as foundational architecture rather than an add-on. The video generation platform implements concrete protections including visible and invisible watermarks on all outputs, C2PA metadata embedding, and internal reverse-image search tools to trace content back to its source. These measures address novel challenges posed by advanced video creation technology.

The platform introduces consent-based controls for real person likeness through image-to-video features requiring user attestation for uploaded photos. Sora Characters provide strong identity control with revocable permissions. Teen accounts receive enhanced filtering, including mature content limitations and parental controls over direct messaging and feed personalization, creating age-appropriate experiences across the platform.

Content filtering employs layered defenses that check prompts and outputs across video frames and audio transcripts to block unsafe material. Audio safeguards prevent imitation of living artists or existing works. Users maintain control over sharing, content removal, and blocking capabilities, with all content subject to reporting mechanisms and clear policy violation recourse. The system represents OpenAI's most comprehensive approach to AI-generated media safety.