thearchivist.funC2PA publishes a first specification for media provenance
The Coalition for Content Provenance and Authenticity released version 1.0 of its technical specification on 26 January 2022. The specification described a way to attach signed provenance information to digital media, including claims about creation and subsequent processing. It was designed as an interoperable technical framework rather than a single platform's private label. A C2PA manifest organizes assertions about an asset and connects them to a signed claim. A validator can check whether the signed information is intact and how it relates to the file being examined. This creates a different kind of evidence from visual inspection. Instead of asking only whether pixels look suspicious, a reader can examine a declared history and the credentials associated with it. Provenance is not the same as truth. A valid signature can support the conclusion that a particular signer made particular assertions and that relevant data has not changed unnoticed. It does not, by itself, establish that a photographed scene was unstaged, that a caption is fair or that a source is trustworthy. The specification's trust model therefore matters as much as the presence of a badge. For an archive, the practical lesson is to preserve provenance alongside the media and explain its scope. Keep the original file when possible, distinguish embedded information from later editorial notes, and record what a validation tool actually reported. A missing credential should not be converted into a claim that a work is fake; older works and unsupported workflows may contain none. The 2022 release marks an attempt to make production history portable across tools. Its usefulness depends on adoption, careful implementation and readers understanding what cryptographic evidence can and cannot show.
Context and provenance
Later entries about TikTok and YouTube show how a technical provenance mechanism can become a visible platform disclosure. These are different layers: a file may contain metadata, a service may interpret that metadata, and an interface may display a simplified label. Each transformation can change what the viewer sees. Researchers should retain the underlying evidence where they can and describe which layer they examined. This also helps avoid treating every AI label as if it came from the same process. Creator disclosure, automated inference and signed provenance can produce similar-looking interface elements while supporting different conclusions.
Why it matters
A common signed-provenance format created a basis for carrying media-production evidence between tools and publishing services.
Sources and evidence
Editorial review: 2026-09-23. New evidence may change this account.