OpenAI adds invisible textGrain watermark to ChatGPT output in the EU
OpenAI announced textGrain, a statistical text watermark for ChatGPT and Codex output, on October 5, 2026. It rolls out to EU users on all plans over the coming weeks, while outside the EU it is API-only and off by default. The detector finds the mark in about 80% of 200-token passages and 95% of 400-token passages, but it collapses under rewriting or translation.
- textGrain is a statistical watermark invisible to users that adds no characters
- EU rollout covers all ChatGPT and Codex plans within weeks
- Outside the EU it is API-only and disabled by default
- Detection drops from 92% to 17% when a quarter of words are swapped
Read next
AI