Watermarking in Generative AI

We study generative AI watermarking, identifying novel vulnerabilities in existing systems and developing new systems for ensuring provenance of generative model outputs, weights, and training data.

Blog Posts

Probing Google DeepMind's SynthID-Text Watermark

We apply the techniques from our recent work to investigate how SynthID-Text, the first large-scale deployment of an LLM watermarking scheme, fares in several adversarial scenarios.

Publications

2025

Discovering Spoofing Attempts on Language Model Watermarks

Thibaud Gloaguen, Nikola Jovanović, Robin Staab, Martin Vechev

ICML 2025

Paper
Code

Watermarking Autoregressive Image Generation

Nikola Jovanović, Ismail Labiad, Tomáš Souček, Martin Vechev, Pierre Fernandez

arXiv 2025

Paper
Code

Robust LLM Fingerprinting via Domain-Specific Watermarks

Thibaud Gloaguen, Robin Staab, Nikola Jovanović and Martin Vechev

arXiv 2025

Paper
Code

Ward: Provable RAG Dataset Inference via LLM Watermarks

Nikola Jovanović, Robin Staab, Maximilian Baader, Martin Vechev

ICLR 2025

Paper
Code

Towards Watermarking of Open-Source LLMs

Thibaud Gloaguen, Nikola Jovanović, Robin Staab, Martin Vechev

WMARK @ ICLR 2025

Paper
Code

Black-Box Detection of Language Model Watermarks

Thibaud Gloaguen, Nikola Jovanović, Robin Staab, Martin Vechev

ICLR 2025

Paper

2024

Watermark Stealing in Large Language Models

Nikola Jovanović, Robin Staab, Martin Vechev

ICML 2024 R2-FM@ICLR24 Oral

Website
Paper
Code