OpenAI Watermarks ChatGPT and Codex: What Changes for EU Users
OpenAI adds invisible watermarks to ChatGPT and Codex output in the EU — here's how to detect them and what it means for compliance.
Why it matters
OpenAI is embedding machine-readable watermarks into ChatGPT and Codex text for EU users, likely in response to AI Act provenance and transparency requirements. Practitioners need to understand detection mechanisms, API implications, and whether watermark stripping affects compliance or output quality.
The key facts
13 to knowWatermarks applied to eligible ChatGPT and Codex text in EU region
Watermarks are invisible (embedded, not visible to end user)
Likely compliance measure tied to EU AI Act provenance/transparency rules
Detection method and limitations not disclosed in article
Scope limited to EU; regional compliance differentiation confirmed
No pricing, API-level detail, or opt-out mechanism disclosed
Invisible watermarks added to ChatGPT and Codex text in EU
Watermarking is selective ('eligible' outputs only—full scope not disclosed)
Detection method exists but article notes 'limits' (specifics not reported)
EU AI Act compliance framing—provenance/attribution requirement
Appears to apply to web interface; API coverage unclear
TechRepublic reporting, not direct OpenAI confirmation of technical details in article
Publication date: October 6, 2026
The story so far
Earlier coverage of this storyline
Go to the source
TechRepublictechrepublic.com
Publisher excerpt: OpenAI will add invisible watermarks to eligible ChatGPT and Codex text in the EU. Learn how detection works, its limits, and what EU users should know.