ChatGPT's EU watermark survives model rewrites, and rewrites silently drop facts
Imaginary_Dinner2710 · reddit · 2026-10-09
After OpenAI began watermarking ChatGPT text for EU users, the author systematically tested the popular 'have another model rewrite it' workaround and found it fails three ways: rewriting models like Claude re-apply their own watermarks (only DeepSeek, GLM, Qwen and local models carry none); rewrites change just 40-73% of words while OpenAI says swapping 25% of words only drops detection from 92% to 17%; and rewrites silently corrupt facts (nearly 3 errors per doc for Qwen3.7 Plus, back-translation through Chinese over 4). He published a free open-source comparison tool using new five-word-sequence ratio as a watermark-survival proxy, plus concrete prompt practices.
More from Models
- OpenAI blocks manual reset with 13% usage left: 'no reset needed right now' — AccomplishedSugar490 · 2026-10-23
- Meta's 816B Muse Spark Open scores 39, losing to 180B Qwen3.8-Flash-Next at 40 — BTA_Labs · 2026-10-23
- Gemini refuses to generate image in Docs, blames document size limit — viksit · 2026-10-23
- Dev jokes his true calling is telling everyone to try the latest DeepSeek — intellectronica · 2026-10-23
- SubQ Snap 1.1 ties for first on Artificial Analysis Decision Index with 384K context — ArtificialAnlys · 2026-10-23
- PolyAI unveils Dialog-RSN-1: audio-native voice model answering in under 300ms — matthen2 · 2026-10-23