Lecture 07 — Chain of custody

Content provenance, the C2PA standard, and what it means to sign a file's history rather than guess at it

5 April 2027

Everything up to this week has been guesswork with increasingly good instruments: infer, from the artefact itself, whether it was likely made or altered by a machine. The Coalition for Content Provenance and Authenticity (C2PA) tries something different — record the truth at the moment of creation, cryptographically, so nobody downstream has to guess. A signed manifest travels with the file: who or what made it, what tools touched it, what was edited and when. It is, deliberately, the same idea as a chain of custody in a criminal investigation: not proof of innocence, but an unbroken, tamper-evident record of who held the evidence and when.

The catch is coverage, not cryptography. A C2PA manifest is only as good as the tools that bother to attach one, and its absence proves nothing — most content in the world today was never signed at all. This week is the hinge of the course: the previous six weeks taught you to interrogate content that carries no record of itself; the rest of the course asks what changes once some content starts to.

Outline#

• how a C2PA manifest is built, signed and read
• what a valid credential can and can’t tell you
• provenance, watermarking and detection: why you eventually need all three, and why none of them is sufficient alone

Reading#

The C2PA Content Credentials explainer (spec.c2pa.org), read as a technical document rather than a marketing one.

Further reading#

• C2PA Content Credentials Explainer — the official specification body’s own explainer, the primary source behind this lecture.
• Content Credentials, explained — a short video walkthrough if you’d rather watch the mechanism than read it.