Developer Claims to Reverse-Engineer Google's AI Watermarking System
By
Jess Weatherbed
A baker's-dozen of insight crammed into one ring.
Summary
A software developer claims to have reverse-engineered Google DeepMind's SynthID AI watermarking system, demonstrating how to strip watermarks from AI-generated images or insert them into other works. The developer, Aloshdenny, open-sourced their work on GitHub, claiming the process required analyzing 200 Gemini-generated images, signal processing, and no proprietary access. Google disputes the claim, stating the reverse-engineering isn't accurate.
Key quotes
· 4 pulledA 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.
No neural networks. No proprietary access,' Aloshdenny said on Medium.
Turns out if you're unemployed and have way too much free time, you can reverse-engineer Google's AI watermarking system.
A claim that, according to Google, isn't true.
You might also wanna read
Reverse Engineering Google's SynthID Watermark: Detection and Removal Through Spectral Analysis
This article describes a GitHub project that reverse-engineers Google's SynthID watermarking system used in Gemini-generated images. The pro
github.com·1mo ago
Google's SynthID: A Watermarking Tool for Identifying AI-Generated Content
SynthID is Google's new watermarking tool designed specifically for AI-generated content. It embeds digital watermarks directly into AI-gene
deepmind.google·3mo ago
Google Introduces SynthID: Watermarking Tool for AI-Generated Content Identification
SynthID is Google's new watermarking tool designed specifically for AI-generated content. It helps users identify whether content has been A
deepmind.google·9mo ago
Remove-AI-Watermarks: Open-source tool strips visible and invisible AI watermarks from generated images
A CLI tool and library called "remove-ai-watermarks" that strips visible watermarks (like Google Gemini's sparkle overlays) and invisible wa
github.com·12d ago
Google's New AI Model Shows Breakthrough Capabilities in Handwriting Recognition and Symbolic Reasoning
Google is testing a mysterious new AI model on its AI Studio platform that appears to be solving two long-standing AI problems: near-perfect
generativehistory.substack.com·6mo ago
Developer Discovers Google Gemini Accessing Personal Context Data
A developer discovers Google Gemini referencing their previous use of the Alembic tool in a response, revealing the AI's ability to access a
unbuffered.stream·6mo ago