OpenAI outlined its approach to text provenance on October 5, in response to EU AI Act requirements. The company says API customers globally can opt in to text watermarking for selected models, with the feature off by default. It also plans to add invisible watermarks to eligible ChatGPT and Codex text output in the European Union over the coming weeks.
OpenAI describes the detector as a limited-access tool for approved researchers and expert organizations. It says the detector will report whether it finds an OpenAI watermark, without exposing a user’s prompts or conversations. That caution is important because the company says text watermarking and detection remain early technologies, with risks of both missed watermarks and false positives.
What a watermark can tell you
A watermark can provide a signal that text may have come from a participating model. It cannot, on its own, prove who wrote the text, whether a person edited it, or whether a passage without a watermark was written by a human. Text can be copied, transformed, translated or generated by a different system. A detector result needs context and should not be treated as a complete authorship verdict.
What this means for teams using AI
For publishers, schools and businesses, provenance is one part of a broader review process. Keep records of how AI assisted with a piece of work, check claims against primary sources, and apply human judgment before publication. If a customer or regulator expects disclosure, follow the relevant policy and law rather than relying on a detector to settle the question.
OpenAI says the EU rollout is regional at launch, not a global default. API customers can choose whether to enable it for eligible models. Teams should therefore check the settings and product terms that apply to their own account. The announcement is a step toward machine-readable provenance, but OpenAI’s own caveats make clear that watermarking will not replace editorial review or transparent workflows.
Sources: OpenAI’s October 5 announcement; the European Commission’s overview of the AI Act.