AI

OpenAI Is Adding Invisible Watermarks to ChatGPT Text in Europe

📅 Oct 6, 2026

✍️ Aamir Raza

OpenAI Is Watermarking ChatGPT Text

A new system called textGrain quietly fingerprints AI-written text — you will never see it.

LexxWire

Europe Gets It First

Eligible ChatGPT and Codex users in the EU will see watermarked text roll out over the coming weeks, per the EU AI Act.

LexxWire

How textGrain Hides the Mark

It plants a statistical signal inside word choices — invisible to readers, readable by a special detector.

LexxWire

Editing Can Erase the Signal

Replacing 10% of words dropped detection from 92% to 66% — and 25% of edits nearly killed it.

LexxWire

The Detector Stays Under Lock

For now, only approved researchers and expert groups can apply to use the detector — it is not public.

LexxWire

More Stories

Reflection AI Unveils Beam, a 501B-Parameter Model Taking on …

Tech Billionaires Made a Record $845 Billion This Year

Read Full Story →

Thanks for reading!

Next: Reflection AI Unveils Beam, a 501B-Parameter Model Taking on China