OpenAI will add invisible watermarks to ChatGPT and Codex text in the EU
OpenAI will begin marking ChatGPT and Codex text in the European Union in the coming weeks. API customers can opt in to watermarking on selected models.
AT A GLANCE
Key points
- Initial ChatGPT and Codex rollout will begin in the EU
- Optional on selected API models; off by default
- The text detector will not initially be public
What is changing?
On October 5, OpenAI announced that eligible ChatGPT and Codex users in the European Union will receive invisible watermarks in text output in the coming weeks. It will not initially be a worldwide default.
The method, called textGrain, creates a detectable statistical pattern in the model’s word choices. Readers cannot see the mark. Optional use is opening worldwide for selected API models; it is turned off by default.
The text detector will initially be available only to approved researchers and expert organizations. Short texts, editing and translation can make detection harder. The absence of a watermark does not prove that a human wrote the text.
The development is relevant to people assessing the origin of AI-assisted text. A watermark alone does not establish a text’s accuracy or how much a person contributed.
Explore the original sources
Source date: 5 October 2026.
Published by AI Radar: 5 Oct 2026, 22:20 (Istanbul, UTC+3). Updated: 6 Oct 2026, 18:44.
FOLLOW THE TOPIC
Topics and tags
Share your discovery.
Send this story to a friend.
Reader comments
What do you think about this development? Share your experience or question. Comments are published after editorial approval.
Your name/nickname and comment become public if approved. Read: Privacy notice.
Loading comments…