OpenAI's Provenance Signals & C2PA

OpenAI utilizes both SynthID watermarks and the C2PA (Coalition for Content Provenance and Authenticity) open technical standard to embed metadata in images and audio, indicating their origin from OpenAI tools.

The top 3

  1. OpenAI's Dual Provenance Strategy: C2PA & SynthID: OpenAI employs a multi-layered approach for content provenance, embedding both C2PA metadata for detailed origin context in DALL-E 3 images and Sora videos, and Google's robust SynthID watermarks for images and audio to ensure persistence even if metadata is stripped.
  2. C2PA Coalition: Who's Building Digital Trust?: The Coalition for Content Provenance and Authenticity (C2PA) is an open technical standard developed by a consortium of industry leaders, including Adobe, Microsoft, Google, Intel, BBC, and most recently OpenAI, aiming to certify the source and history of digital content to combat misinformation.
  3. Digital Provenance Limits: Can C2PA Watermarks Be Bypassed?: While C2PA uses cryptographic signatures to detect tampering with embedded manifests, it doesn't prevent complete removal of provenance data, for instance, by saving an image without its manifest or taking a screenshot. This highlights a key limitation, though combining C2PA with resilient watermarks like SynthID aims to mitigate such 'strip attacks'.

Sources

Open the full topic