Content Credentials (C2PA) explained for fraud teams

Content Credentials are a signed record, built on the open C2PA standard, that can travel with a file and say which device or app made it and what was done to it since. A valid credential shows that the file hasn't changed since it was signed, and who signed it. It doesn't show that what the file depicts is true. A missing credential proves nothing either, because most files are made without one and many platforms strip them. For fraud teams, a credential is useful evidence when it's there, never a pass or a fail on its own.

What a Content Credential contains

The C2PA specification describes three parts:

• Assertions. Statements about the file, such as the camera or app that created it, the edits made, the source files used and whether AI tools were involved.
• A claim. A digitally signed, tamper-evident bundle of those assertions, signed with the signer's X.509 certificate. The signer is often the maker of the camera or app.
• A hard binding. A cryptographic hash of the content. If the file changes after signing, the hash no longer matches and validation fails.

What a valid credential tells you

• The file hasn't been modified since the credential was signed.
• Which device or software signed it, as named in its certificate.
• The history the signing tools recorded, including AI generation when the tool declares it. OpenAI, for example, added C2PA credentials to DALL-E 3 images made in ChatGPT and its API.

A credential that fails validation is a strong lead. It can mean the file was changed after it was signed.

What it doesn't tell you

• Whether the content is true. The C2PA's own explainer says provenance information alone "cannot tell you whether the digital content is true, accurate or factual." A camera can sign a genuine photo of a staged scene, or of a screen.
• Whether the signer can be trusted. In 2025, a researcher showed that a multiple exposure feature on Nikon's Z6III could be used to bypass its authenticity check, and Nikon said it would revoke the C2PA certificates it had issued.
• What happened before signing, or in tools that don't record their steps.
• Who is in the file, or who sent it.

Why a missing credential proves nothing

• Most files are made without one. Google says the Pixel 10 is the first phone lineup with Content Credentials built into every photo from its camera app. Samsung's Galaxy S25 adds them to content created or edited with generative AI. Some professional cameras sign images, such as the Leica M11-P, and Sony has added C2PA signing for video on some cameras.
• Platforms strip them. The C2PA's guidance for implementers says metadata, including credentials, may be routinely removed by platforms that don't support C2PA, and that this is common on social media. A screenshot drops them too.
• They can be removed on purpose. The explainer says so directly.

Asked whether to distrust media without Content Credentials, the C2PA's explainer answers "Maybe," and adds that no assumption should be made about a file's trustworthiness purely from whether it uses them. In practice, a missing credential means you check the file itself.

To make credentials harder to lose, the C2PA supports durable credentials: invisible watermarks or fingerprints that can help find a credential after the metadata is stripped.

How to check a file by hand

1. Open the file in a credential viewer, such as the Content Authenticity Initiative's Verify site. Check your data handling rules before you upload a customer's file to any outside site.
2. Read the result. Valid: note who signed it and what the history says. Invalid: the file may have changed after signing, so look closer. None: no conclusion either way.
3. Compare the history with the story. A claim photo said to be taken at the scene whose credential names an AI tool, or a selfie whose credential names an editing app, deserves a question.
4. Check the file itself either way: edits, generation traces, metadata and whether it fits the rest of the case.

Where software helps

Software can read and validate credentials on every file at intake, check the signatures and certificates, and flag credentials that fail. Just as important, most files arrive with no credential at all. Software can check those for edits, AI-generation traces and metadata gaps, which is where most of the evidence still is.

FAQ

Do Content Credentials prove a photo is genuine?

No. A valid credential shows the file hasn't changed since it was signed, and who signed it. It doesn't prove that what the photo shows happened.

If a file has no Content Credentials, is it fake?

No. Many cameras, phones and apps don't add them yet, and many platforms remove them. Check the file itself.

Can Content Credentials be removed or faked?

Removed, yes: stripping metadata or taking a screenshot removes them. A forged or altered credential should fail validation, unless the device or software that signs it can be tricked, as the Nikon case showed.

Does a credential that says AI-generated mean fraud?

No. An AI-made image can be legitimate, such as a product render or an illustration. What matters is whether it is being presented as a real photo of a real event.

Are Content Credentials the same as a watermark?

No. A credential is signed metadata attached to the file. Watermarks and fingerprints are what the C2PA uses to help recover a credential after that metadata is lost.

How Trustable helps

Reading Content Credentials in Trustable is in development. Today, Trustable checks the file itself: documents, images, video and IDs for edits, AI generation and metadata, with each finding marked where it sits and how strong it is, for your team to review. That covers the many files that arrive without a credential. See content authenticity for documents, images and video, or Book a demo.