OpenAI announced text watermarking on 5 October. API customers worldwide can opt in for selected models; eligible ChatGPT and Codex text in the EU is scheduled to receive watermarks over the coming weeks.
Based on 2 primary sources · See sources
An origin signal with limits
The textGrain system adds a statistical signal through word choice. Short passages and later editing can make detection less reliable. Initial detector access is restricted to approved researchers and expert organisations.
OpenAI explains its phased text provenance approach.
View the original post on X ↗What a result cannot establish
OpenAI says detection cannot establish accuracy, ownership, user identity or the extent of human contribution. A missing signal does not prove human authorship.
For marketers
Our reading: marketing teams should preserve their own drafting and approval records. A detection result alone is a poor basis for assessing a writer’s work or clearing a campaign claim.
Keep source checking, editorial approval and provenance records as separate steps.
Sources
- EU text provenance announcement
OpenAI · Announcement ·
- Original text provenance announcement on X
OpenAI · Announcement ·
Written with AI assistance from the sources above. The marketing implications are our interpretation. Our editorial approach.
