Our approach to EU text provenance rules
If you or your EU-based clients generate content with ChatGPT or Codex, outputs will soon carry an invisible provenance watermark as part of OpenAI's EU AI Act compliance work — this happens automatically and needs no action from most users. If you build on the API, you can already opt in to watermarking.
What changed
Source summary
In response to EU AI Act requirements, OpenAI is rolling out "textGrain" watermarking for AI-generated text: API customers can opt in to text watermarking immediately, ChatGPT and Codex users in the EU will have invisible watermarks added to their outputs in the coming weeks, and researchers/organisations can apply for access to the watermark detector. OpenAI notes the detector has non-trivial false positive/negative rates and weakens when text is edited, which is why broader public detector access is not yet available.
Why it matters to your work
Practical interpretation
If you or your EU-based clients generate content with ChatGPT or Codex, outputs will soon carry an invisible provenance watermark as part of OpenAI's EU AI Act compliance work — this happens automatically and needs no action from most users. If you build on the API, you can already opt in to watermarking.
What to try next
Mention to EU-based clients or stakeholders that AI-generated text from ChatGPT/Codex will soon carry an invisible provenance watermark — useful context if anyone asks about proving content authenticity.
Source & availability
This is a summary of a published provider announcement, with practical context from SCALA’s feed.
Plan, region, workspace and rollout conditions may affect access in Australia. Check the original source for current availability.
openai.com