Educational simulation · Content provenance
The Watermark Stress Test is an interactive, in-browser lab that shows why AI content watermarks — the steganographic approach used by systems like Google SynthID — are fragile against ordinary edits, and why durable provenance needs cryptographically signed Content Credentials (C2PA).
It injects its own synthetic watermark and analyses it entirely in your browser. It does not detect, contain, or remove Google SynthID or any production watermark, and nothing you load is uploaded.
Six hands-on labs — Image, Text, Audio, Analysis, Credentials, and a live Perturbation Pipeline — plus a full glossary and the case for C2PA. Read the companion white paper: Why AI watermarks break →
Loading the interactive lab…