C2PA explained

What is C2PA, and what are Content Credentials?

C2PA is an open technical standard for attaching tamper-evident provenance information to digital media. Content Credentials are a way people can inspect that signed evidence.

C2PA helps answer questions about origin and declared history. It is not a universal detector for truth, falsity or AI generation.

The credential

Signed evidence travels with the media

A Content Credential combines structured declarations with cryptographic validation.

ManifestThe signed container

Holds the claim and assertions associated with the asset.

AssertionsThe declared information

Can describe actions, ingredients, software and other provenance context.

SignatureThe integrity mechanism

Binds the signed claim so changes can be detected during validation.

CertificateThe signer context

Connects the signature to an identity and a trust-chain evaluation.

Lifecycle

From creation to verification

The standard supports a chain of signed declarations rather than one permanent truth label.

01Create or edit

A compatible tool produces or transforms an asset.

02Declare

The tool records available provenance assertions.

03Sign

A certificate and private key sign the claim.

04Verify

A validator checks integrity, signature and trust context.

Right question, right tool

C2PA is different from detection

Provenance evidence and content classifiers can complement each other, but they measure different things.

Verifies available evidence

  • Is a signed claim present?
  • Does it still match the asset?
  • What history was declared?
  • How is the signer evaluated?

Detection estimates content properties

  • Does a model classify the content as synthetic?
  • How confident is that probabilistic estimate?
  • Could processing affect the score?
  • What false-positive rate applies?
Core limitation

Valid provenance does not guarantee factual truth

A credential can show that an identified signer made a declaration and that the signed evidence validates. Reviewers still need editorial, legal or domain-specific methods to assess the underlying claims.