Intelligence · Explainer
Content provenance standards are shipping. Almost nothing checks them.
Cameras, editing software and several large platforms now attach or preserve provenance metadata. The verification layer that would make it useful to a reader barely exists.
Independent coverage
Published 5 September 2026
6 min read
Evidence: Reporting
The technical work on content provenance has gone better than most observers expected. Signing at capture, preserving edit history, and declaring synthetic generation are all specified and increasingly implemented.
The deployment gap is on the reading side. A signed image posted to a platform that strips metadata is an unsigned image. A signed image displayed in an interface with no verification affordance is an unsigned image to the person looking at it.
What the standard does
It creates a tamper-evident record of capture device, edits and generation claims, cryptographically bound to the file.
What it does not do
It does not establish that the scene is genuine. A photograph of a staged event carries perfect provenance. This distinction matters because the public framing has repeatedly promised authenticity when the technology provides custody.
Where it will be genuinely useful first
Not in social feeds. In insurance claims, court evidence, journalism workflows and regulated inspection records, where the chain of custody question is the actual question and the verifying party is motivated and technical.
What would change consumer usefulness
Metadata preservation as a platform obligation rather than a choice. Until upload pipelines stop stripping it, the reader-facing case remains theoretical.
"Provenance proves where a file came from. It cannot prove that what the file shows is true, and the two keep being confused."
Sources