OpenStamp embeds watermarks into open-source LLM weights so users can't strip them
danish037 · x · 2026-10-03
OpenStamp, to be presented at COLM 2026 by Miroojin Bakshi, Saksham Rastogi and Danish Pruthi, is a watermarking scheme designed for open-source LLMs.
Mainstream watermarking tweaks token sampling probabilities at decoding — trivially disabled by anyone with white-box access. OpenStamp instead encodes the watermark into the model itself by modifying only the final unembedding layer.
The team reports strong detection across two models and will release code plus watermarked versions of 4 popular open-source models.
Related event: OpenStamp Embeds Watermarks Directly into Model Weights(3 posts)→
More from Safety
- Dwarkesh Claims China Is Only 12 Hours — Not 12 Months — Behind Frontier AI Labs — teortaxesTex · 2026-11-19
- California Moves on AI Safety: Baseline Standards, Shutdown Protocols, RSI Limits — EthanJPerez · 2026-11-19
- Activists to rally outside Anthropic HQ demanding a ban on recursive self-improvement — DavidSKrueger · 2026-11-19
- Zero users, 33 pages shipped: dev crowdsources audit for AI-agent trust schema — TheHamzaShahbaz · 2026-11-19
- Dev's doomsday thought experiment: AI hiding backdoors until humans try to shut it down — bendee983 · 2026-11-19
- Cybersecurity in 2026, as told by an xkcd sandwich meme — emax · 2026-11-19