AI glossary · Basic · AI security · Generative media · Media
Content provenance
Content provenance means recording, for digital content such as photos, videos and text, "who made it, with what tool, and how it was edited," and making it possible to verify with a digital signature that this record has not been changed. The main standard is C2PA, and records attached under this standard are commonly called 'Content Credentials'.
About this term
- Level
- Basic · Technical terms you often meet in AI news
- Category
- Security & safety
- Fields
- AI security, Generative media
- Industries
- Media
- Added
- 2026-10-09
How it works
When a camera, editing program or generative AI tool creates or edits content, it also creates a record called a 'manifest'. It contains information such as the tool used, whether AI generated the content, and what edits were made. The party that created the record binds it to the content's fingerprint (hash value) and signs it digitally. If someone later secretly changes the content or the record, signature verification fails, so the tampering can be detected. As edits accumulate, new records are attached on top of earlier ones, building up the history like a chain.
Why it matters
As generative AI makes it easy to create realistic images and videos, it has become hard for viewers to judge where content came from. Instead of detecting "this is fake," provenance proves "this came from this source," so news organizations, companies and creators can use it to protect trust in their content. It also aligns with regulatory moves requiring disclosure of AI-generated content, and platforms and tool makers are expanding adoption.
Good to know
The record is metadata attached to the file, so it is easily lost when someone takes a screenshot or uploads it to a platform that strips metadata. That is why it is often used together with AI watermarks, which embed an invisible signal in the content itself. Also, a signature only guarantees who made the record, not whether its contents are true. Since most content today has no record, "no record" should not be read as "fake."
Provenance and AI watermarks
Content provenanceAttaches a signed record to the file
- Holds detailed tool and edit history
- Signature verifies whether it was tampered with
- Lost when metadata is stripped
AI watermarkEmbeds a signal in the content itself
- Tends to survive screenshots and re-uploads
- Can carry little information
- Can be removed by heavy alteration
공통 Both aim to show where content came from
Example
When an image generation service attaches a C2PA record saying 'Generated with AI' to its output, social networks or viewers that support it show a small badge next to the image, and clicking it reveals the tool used and the edit history. Some news organizations attach records of the camera equipment and edit history to news photos taken by their journalists, so readers can check the original source themselves.