5 October 2026
OpenAI adds invisible watermarks to ChatGPT text in EU, optional for API users
First reported
TechCrunch, The Decoder and 1 other ran this on , all on the same day.
- OpenAI is rolling out textGrain, an invisible statistical watermark embedded in ChatGPT and Codex word choices, to EU users over coming weeks to comply with the EU AI Act's requirement that AI-generated text be machine-readable.
- API customers worldwide can opt in to watermarking starting today, unlike Anthropic which makes Claude's watermarks mandatory globally. Detector access is restricted to approved researchers and organizations initially.
- OpenAI's tests show the watermark is detectable in about 95 percent of longer passages but drops to 80 percent in shorter text and fails on math content. Editing just 10 percent of words reduces detection to 66 percent.
Our read
Earlier on this story
Written by Newskeryx from which newsletters covered this story and from our own archive.
Reported by The Decoder, The Verge, TechCrunch