Content Credentials are evidence about a file, not a verdict on its story

A practical distinction between provenance, factual accuracy and permission, with a handoff checklist for publishers.

Conceptual illustration of connected translucent cubes representing a chain of media records.

A signed history can answer an important question about an image: what information accompanied this asset, and has that information remained bound to it? It cannot settle every question a publisher needs to ask. A photograph may have a well-documented origin and still carry a misleading caption. A generated illustration may be openly described and perfectly appropriate to its context.

The useful editorial distinction is between the history of the file and the truth of the claim made with it. Treating them as one question turns a technical signal into a promise it was never designed to make.

What the credential contributes

C2PA’s version 2.3 explainer describes Content Credentials as cryptographically bound provenance structures containing assertions about an asset. Validation can assess their association with the asset, integrity and trust status. The explainer explicitly separates that process from deciding whether the provenance information is true. C2PA’s goals and non-goals.

For an editor, that makes a credential something to read rather than a badge to admire. Which actions are described? What part of the chain is available? What does the validator actually say? Preserve the result with the file version you inspected. A note saying “has credentials” is too imprecise to survive a later export or dispute.

Three questions that must remain separate

Provenance: what recorded history travels with the asset? Accuracy: does the published image, caption and context support the story being told? Permission: do you have the necessary authority to use the material in this way? A useful review asks all three and assigns an owner to each.

Imagine a genuine photograph from an earlier flood reused beside a report of a new one. The file history may be intact; the caption can still mislead. Now imagine an openly labelled illustration of a hypothetical future station. It may be entirely synthetic while serving its stated explanatory purpose. These are illustrative scenarios, not incidents investigated by this journal. They show why a single “real or fake” field is too crude.

Absence is an incomplete observation

C2PA’s FAQ explains that provenance metadata can be separated from an asset, and that supported soft-binding approaches can help rediscover associated credentials. It also distinguishes provenance from digital rights management. A credential is not itself a mechanism that grants or restricts use. C2PA FAQ on stripped metadata and DRM.

Our editorial inference is therefore narrow: when no credential is available to a particular check, record that absence. Do not convert it into a claim that the image is fabricated. Equally, do not treat successful validation as evidence that every ingredient and every contextual assertion has been independently investigated.

The question becomes practical: can the supplier provide the source file, the relevant creation record and an explanation of material edits? If the story depends on knowing those things and they remain unavailable, the publication decision should reflect the gap.

Make the handoff carry the evidence

A useful asset handoff has five parts. Keep the approved file, its stable identifier or hash, a readable record of the provenance check, the editorial caption and the permission record. If any part changes, state which one. A fresh caption does not automatically require a new image, but it does require reconsidering what the image appears to establish.

Test the delivery path too. The asset in a production folder and the version served by a publishing platform are separate objects. Inspect the actual published derivative when the workflow depends on a credential being available. Preserve the source even when the delivery system changes format or resolution.

Do not expose private supplier or contributor information simply to make the record look complete. Decide which evidence belongs in an internal log and which explanation belongs beside the image. Readers usually need a clear description of what was created or changed; an editor may need additional restricted records to support that description.

A more useful publication note

Prefer a statement about the actual intervention: “The background was replaced,” “This is an illustration of a proposed design,” or “The speaker’s translated voice was synthesised with permission.” Use such wording only when supported by the production record. It tells the reader how to interpret the material.

A provenance system can strengthen that record. It cannot take responsibility for the sentence around the picture. The editor’s task remains to connect the available evidence to a claim of the right size, then preserve enough information that another person can check the decision.