C2PA

How it works

C2PA (Coalition for Content Provenance and Authenticity) specifies Content Credentials: a cryptographically signed manifest bound to a media asset that records where the asset came from and what was done to it. The manifest holds assertions about origin, edits, the tools involved, and whether generative AI was used. Verification recomputes the asset hash, checks it against the manifest, and validates the signing certificate. Each conforming edit adds a new manifest referencing the previous one, forming a provenance chain. The specification moved through version 2.2 in May 2025, 2.3 in January 2026, and 2.4 in April 2026, with a conformance program for interoperability across creation tools, publishers, and verifiers.

When it matters

It matters for publishers, newsrooms, and anyone whose original media would be damaged by an alteration claim, and increasingly for AI-generated content facing disclosure requirements. The practical difficulty is not signing but survival: most image pipelines strip metadata by default, so a manifest rarely reaches the delivered asset without a pipeline audit. See Implement Content Credentials.

Common misconceptions

• "A valid credential means the claims are true." It means an identified signer made those claims about these exact bytes. A dishonest signer produces a valid credential containing a lie.
• "It detects AI-generated content." It does not. A generator that does not sign produces an unsigned file, indistinguishable from an unsigned camera photo.
• "Absent credentials are suspicious." Nearly all media has none. Presence is informative; absence is not, and a UI that flags absence trains users to ignore the indicator.