What is C2PA?

C2PA (the Coalition for Content Provenance and Authenticity) is an open technical standard for attaching verifiable provenance to digital content — cryptographically signed metadata that records where a file came from, what made it, and whether it has been altered since. The consumer-facing name for a C2PA record is a Content Credential.

In one paragraph: when an image (or video, or audio file) is created or edited, C2PA-capable software embeds a signed manifest inside the file. Anyone — with any compatible validator, no account, no permission — can then check who signed it, when, what tools were involved, and whether a single pixel has changed since. It is the closest thing digital content has to a tamper-evident seal.

How the C2PA standard works

A C2PA manifest is embedded in the asset itself (in a JUMBF container) and has four moving parts:

Assertions — machine-readable statements about the content. The one that matters for AI is the provenance action carrying an IPTC digitalSourceType such as trainedAlgorithmicMedia ("this was generated by a trained model").
A hard binding — a cryptographic hash over the content itself, so any pixel-level edit breaks validation visibly.
A claim — the structure tying assertions, binding, and the generator's identity together.
A digital signature over the claim, carrying the signer's certificate chain and (in good implementations) an RFC 3161 trusted timestamp proving when it was signed.

Trust is layered. Any validator can confirm a signature is structurally valid; whether the signer is recognised depends on their certificate chaining to the official C2PA Trust List, which requires completing the C2PA Conformance Program. That distinction — valid vs. trusted — is central to reading any verification result honestly.

What are C2PA metadata and Content Credentials?

Same technology, different vantage points. "C2PA metadata" is what developers call the embedded manifest; "Content Credentials" is the consumer-facing label (and the "Cr" pin) used when platforms display it. LinkedIn, TikTok, and camera and phone makers have begun signing or surfacing Content Credentials; Adobe's public verify page and the open-source c2patool read them. The standard's reference implementation (c2pa-rs) is open source with bindings for Node and Python.

C2PA vs SynthID (and watermarks generally)

The short answer: different layers, not rivals. C2PA is a public, verifiable record — anyone can check it, it names the signer, and it proves integrity — but it lives in metadata, which pipelines can strip. SynthID and similar imperceptible watermarks survive many pixel-level transformations, but detection is proprietary and the attribution is coarse. Robust Article 50 implementations treat signed metadata as the verifiable record and a watermark as the survivability layer — the developer guide covers the combination in depth.

Why C2PA matters right now: the EU AI Act

Article 50 of the EU AI Act requires providers of generative AI systems to mark outputs as AI-generated in a machine-readable form. Under the AI Act as currently applicable, that obligation applies from 2 August 2026 to all in-scope systems, regardless of when they were first placed on the EU market. (If the agreed Omnibus grandfathering provision is formally adopted, qualifying systems already on the market before that date would have until 2 December 2026 for the Article 50(2) marking requirement — pending final publication in the Official Journal, not yet law.) The EU's Transparency Code of Practice names C2PA-style signed metadata as a reference marking measure. Whether the obligation applies to your system depends on your circumstances — the scope screener gives an orientation, not a determination.

See it, then ship it

Check any file right now — drop an image on the free verify page and read its Content Credential (or absence).
Add C2PA marking to your product — one call in your generation pipeline; media bytes never leave your process:

npm install @oprindo/sdk   # or: pip install oprindo

const result = await oprindo.mark(
  { buffer, mimeType: "image/jpeg" },
  { generator: { name: "YourProduct", version: "1.0.0" } },
);

The Article 50 developer guide walks the full implementation — manifests, the AI-generated assertion, watermarking, detection, and the evidence records that let you show your work. Every claim this platform makes about its own signing is backed by artifacts you can check yourself on the proof & security page.