Has Google’s AI watermarking system been reverse-engineered?
TL;DR
A software developer claims to have reverse-engineered Google DeepMind's SynthID system, showing how AI watermarks can be stripped from generated images or manually inserted into other works. A claim that, according to Google, isn't true. The developer, going by the username Aloshdenny, has open-sourced their work on GitHub and documented his process, claiming all it required was 200 Gemini-generated images, signal processing, and "way too much free time. " A little weed also seemed to help.
Nauti's Take
The upside: public security research, real or not, accelerates better solutions. The risk: Google's rebuttal stays vague, which isn't reassuring.
If SynthID is genuinely vulnerable, AI watermarks need this setback as a wake-up call to become more robust.
Sources