OpenAI Details Sora 2 Safety And Launch Path
System card cites high block rates, C2PA metadata, and moving watermarks as the model reaches web and iOS with guarded uploads.
What OpenAI published
OpenAI released the Sora 2 system card on September 30, outlining safety design, evaluations, and deployment scope. The video and audio generator operates under AI Safety Level 3 protections. Sora 2 is available via sora.com and an iOS app, with API access planned. :contentReference[oaicite:6]{index=6}
Safety results and controls
Evaluations showed 96.04 percent blocking for adult nudity without likeness and 99.70 percent for self‑harm content, with overrefusals below 6 percent across categories. The stack includes multimodal moderation, a dedicated child safety layer, and restrictions on photorealistic person uploads. Generated assets carry C2PA metadata and visible moving …
Archive Access
This article is older than 24 hours. Create a free account to access our 7-day archive.