Prevention, not detection
Proof at the moment of capture. Not suspicion afterwards.
There are three ways to deal with a doubtful image: sign the file, detect the fake, or prove the capture. Only the third one ages well.
| Approach | What it establishes | What it cannot establish | How it ages |
|---|---|---|---|
| Signature content credentials | The file is unchanged since it was signed. | Whether the lens saw reality or a screen, a print, a generated image [2]. | Stable – but silent on the scene. |
| Detection forensic and AI models | A probability that an image was generated or edited. | Certainty. A verdict a court can rely on. Anything about the moment or presence. | Loses ground with every new generation of image models. |
| Proof of real capture VeraCapt | Real scene, real moment, presence, completeness – at a stated level. | The truth of what is shown; anything about files not captured with it. | Does not age: it is evidence, not a guess. |
Sources
- C2PA Technical Specification 2.x, section 4.2.2.8: a valid manifest and signature do not prove that the captured content depicts a real scene; the specification names the recapture of a screen or print as out of scope.
Economics
Detection scales with fraud. Prevention scales with captures.
A detection pipeline costs more the more suspicious files it sees, and is never finished. A proof at capture costs the same for every capture, and removes the file from the pipeline altogether.
VeraCapt is not a replacement for your review process. It is the reason review does not have to run on the captures that carry a proof.