Generative AI systems can now produce photorealistic images, synthetic audio, and convincing video at scale. This capability makes it increasingly important for anyone consuming digital media to be able to understand the extent to which the asset was created or modified by AI. To address the idea of content provenance, the C2PA publishes Use of Content Credentials to Identify Synthetic and Non-Synthetic Content, a white paper that clarifies how developers, creators, and platforms can use C2PA’s Content Credentials to communicate that information clearly and precisely.
This paper focuses specifically on how Content Credentials can be used to label AI-generated and AI-modified content, as well as non-synthetic content, recorded within tamper-evident, cryptographically signed manifests. Publishing this paper continues the C2PA’s deliberate effort to make the open standard both more robust and easier to implement.
The paper addresses five core areas:
- A machine-readable classification for AI content – The digitalSourceType field, drawn from the IPTC vocabulary, provides the primary signal for declaring whether an asset was generated entirely by a model, enhanced by AI, captured by a device, or edited by a human.
- A dedicated AI Disclosure assertion – The new c2pa.ai-disclosure assertion captures model provenance, scientific domain, and the degree of human oversight, offering transparency well beyond a binary “AI-generated” label.
- Precise localization of AI modifications – Using Regions of Interest, Content Credentials can identify exactly which portions of an asset were affected by AI: a region of an image, specific paragraphs in a document, or defined segments of audio and video.
- Documentation of the generation recipe – Through inputTo ingredient assertions, the prompts, reference images, seed values, and parameters behind a generation can be recorded as verifiable provenance data.
- A standardized vocabulary for creation and editing – Through the Actions assertion, a Content Credential can record each step in an asset’s lifecycle using consistent terms such as c2pa.created, c2pa.opened, and c2pa.edited. Because an AI usage indicator can be applied to each individual action, a single file can distinguish a human-created original from a subsequent AI-assisted edit, giving audiences a precise account of when and how AI was involved.
As AI-generated and modified content becomes more prevalent, Content Credentials are a dependable foundation for establishing authenticity across every media type. With more than 500 members and over 6,000 affiliates now supporting the standard, the C2PA continues to evolve the Content Credentials specification to keep pace with the dynamic content provenance landscape. The C2PA encourages the technical community to use this publication as a way to provide clarity and standardization of terms for content provenance.