C2PA

C2PA and Content Credentials, explained plainly

2 min read

C2PA is an open standard for attaching a signed history to a file: which app made it, which edits followed and whether generative AI was involved. Adobe, Microsoft, Google, OpenAI and several camera makers write it, and social platforms read it.

AI Label Remover shows the claim generator, each recorded c2pa action and any DigitalSourceType value it finds. Washing removes the entire manifest store from the file.

Wash a file now — it stays on your device

Drag and drop your images here

JPG, PNG, WebP, GIF, HEIC, AVIF · each file up to 25.00 MB and 40 MP · 50 files per session

The short version

The Coalition for Content Provenance and Authenticity (C2PA) maintains an open standard for attaching a tamper-evident history to media. Users see it as Content Credentials. Cameras, editors and AI generators can sign a record into an image saying how it came to be — shot on a camera, edited in an app, or produced by a model.

Anatomy of a manifest

A manifest is a compact, structured bundle with three main parts:

  • Claim generator — the software or service that produced the manifest, such as an editor or an AI image API.
  • Assertions — statements about the content: c2pa.actions (created, edited, opened, resized and so on), a digitalSourceType and links to ingredient files.
  • Signature — a cryptographic signature and certificate identifying the signer, plus a hash binding the manifest to the exact image data.

Where each format keeps it

The manifest is packed in a JUMBF container that each format stores in its own slot: APP11 segments in JPEG, a caBX chunk in PNG, a C2PA chunk in WebP and a uuid box in HEIC and AVIF. As a separate block it can be read and removed without decoding the image. AI Label Remover shows the block, the claim generator, the recorded actions and the source type whenever they exist.

What removal actually does

Washing deletes the embedded record from that copy of the file. Bear in mind that any edit already breaks the cryptographic binding — verifiers report a manifest that no longer matches its pixels as invalid — so stripping removes a record that would otherwise ride along.

Some systems add “durable” credentials: an invisible watermark or fingerprint that lets a verifier fetch the manifest from an online repository even after the embedded copy is gone. Metadata washing has no effect on those, because they live in the pixels or on a remote server.

Keep it, or wash it?

Credentials are useful when you need to show where an image came from — in reporting, or for your own original photography. They can also disclose software, edit history and identity you would rather keep to yourself, or set off AI labels on lightly edited photos. Read the manifest first and decide per file.

About C2PA

Who uses C2PA?

Adobe, Microsoft, Google, OpenAI, Meta, camera makers such as Leica, Sony and Nikon, and many news organisations support the standard or belong to the coalition, though each uses it differently.

Is C2PA a kind of invisible watermark?

No. C2PA is metadata stored beside the image data. Pixel watermarks like SynthID are a separate mechanism and are unaffected by washing.

How do I verify the signature before washing?

AI Label Remover shows generator, actions and source type. For full signature validation, run the original through an official Content Credentials verifier first.

Does removing C2PA cost any quality?

No. The manifest is its own block; the compressed image data is copied as-is.

All articles