Creating with Sora Safely

Source: OpenAI Date: 2026-03-23 URL: https://openai.com/index/creating-with-sora-safely

Summary

Title-only: A post from OpenAI on safety considerations for creating content with Sora — their video generation model. March 2026 is well into Sora’s commercial life (launched publicly December 2024), and this post likely covers updates to safety guidelines, content policies, and technical measures for preventing misuse (deepfakes, synthetic media for disinformation, non-consensual intimate content). The framing as “creating with Sora safely” is creator-oriented, not just policy-oriented.

Implications

The synthetic video safety thread. Sora’s video generation capability creates entirely new categories of risk: convincing deepfake videos, synthetic news footage, and realistic-but-false documentary content. “Creating with Sora safely” as a creator-facing guide suggests OpenAI is building safety culture among Sora users — similar to how platforms like YouTube have creator guidelines for responsible content. The challenge is that unlike text, synthetic video is harder for viewers to identify as AI-generated.

Content provenance. The Sora safety post likely discusses C2PA (Coalition for Content Provenance and Authenticity) metadata standards that cryptographically authenticate the origin of digital content. OpenAI joining the C2PA working group and implementing provenance metadata in Sora outputs is the technical mechanism that “creating with Sora safely” depends on — without provenance marking, safety guidelines alone cannot prevent misuse of AI-generated video at scale.