MetaTaken

C2PA / Content Credentials

Remove supported C2PA Content Credentials — and verify what remains.

Content Credentials are provenance data associated with the C2PA standard. MetaTaken can inspect supported embedded C2PA / JUMBF structures, remove them with Full Metadata Clean where it is safe to do so, and re-inspect the output.

Download MetaTaken for Windows See how MetaTaken works
Local Windows processing No image uploads No account Automatic verification

What are C2PA Content Credentials?

C2PA is a provenance standard for attaching verifiable information about a digital asset's origin and history. In the C2PA specification, a Content Credential is the non-technical term for a C2PA Manifest; manifest stores use JUMBF-compatible structures and can carry signed claims and assertions about an asset.

For an ordinary user, the practical point is simpler: an image can carry machine-readable provenance that says something about how it was created or edited.

MetaTaken treats C2PA as data to inspect, not a magic “AI yes/no” flag.

Content Credentials may contain useful provenance and are not inherently “bad metadata”. But there are legitimate reasons to remove embedded provenance from a delivery copy — privacy, workflow separation, client requirements, or simply wanting to choose what metadata leaves your computer.

MetaTaken therefore shows the finding first. Full Metadata Clean can target supported embedded C2PA / Content Credentials / JUMBF structures along with other safely removable metadata. The original image is not overwritten.

How to remove C2PA / Content Credentials with MetaTaken

  1. Inspect the image. MetaTaken identifies supported provenance structures and other metadata before cleaning.
  2. Choose Full Metadata Clean. This is the mode that targets supported C2PA / Content Credentials / JUMBF provenance.
  3. Keep the original. MetaTaken writes a separate cleaned copy rather than cleaning the source in place.
  4. Verify the output. The cleaned file is inspected again and the result is reported as Clean, Review, Failed or another evidence-based state.

Removing Content Credentials is not the same as removing every AI signal.

C2PA is a file-level provenance mechanism. Pixel-level watermarking is a different mechanism, and a platform can also run independent classifiers. MetaTaken does not claim that removing C2PA makes an image universally “not AI”.

Likewise, because C2PA can support external provenance models as well as embedded data, MetaTaken reports what it actually detects in the file and what it can safely remove rather than making a blanket claim about every possible provenance source.

Why verification is part of the feature

A cleaner should not infer success from “the command completed”. MetaTaken re-inspects the cleaned copy. Where supported, it also compares decoded pixel content before and after and only reports Pixel content unchanged when that check actually passes.

MetaTaken's rule: detected, removed, retained, unsupported, uncertain and not tested are different states. The report should say which one applies.

Related MetaTaken guides

Remove AI metadataClean supported generator and provenance metadata from a separate copy. AI detection vs metadataWhy C2PA removal is only one layer of a wider detection question. Process locally on WindowsNo MetaTaken cloud image-processing service receives your image.

Inspect. Clean. Verify.

See what the file is actually carrying.

MetaTaken is a free local Windows utility. It inspects supported image metadata, creates a separate cleaned copy, then inspects the result again instead of assuming the job worked.

Download MetaTaken v0.3.0-rc2 Release notes and checksum on GitHub