Published : 06 Oct 2026, 02:57 PM
Updated : 06 Oct 2026, 02:57 PM
OpenAI is adding an invisible watermark to eligible ChatGPT and Codex text output in the European Union over the coming weeks, in response to the EU AI Act.
The Act requires generative AI providers to make generated text identifiable in a machine-readable way.
The company said in an announcement that text watermarking and detection remain early technologies with significant limitations, so it is stressing transparency about what a watermark can and cannot show.
The rollout will cover all plans in the EU only, not as a global default. From Thursday, API customers worldwide can opt in for select models, with the feature off by default.
The company is also opening applications for its detector, but only to approved researchers and expert organisations at first.
The tool will report whether it finds an OpenAI watermark, without identifying the user or revealing prompts, and will not be public at launch because of the risk of missed watermarks and false positives.
The technology, called textGrain, adds an invisible statistical signal to the model's word choices.
OpenAI says it matched or exceeded other approaches it tested, including SynthID for text, though strong performance under ideal conditions does not guarantee reliable everyday detection.
At a 1 percent false positive rate, the detector found watermarks in about 80 percent of 200-token passages and about 95 percent of 400-token passages on topics such as psychology, but far fewer in maths.
Replacing 10 percent of words with synonyms cut detection in 400-token passages from about 92 percent to 66 percent, and replacing 25 percent cut it to 17 percent.
A watermark does not measure human contribution, prove ownership, identify the user or verify accuracy, the company said, and a missing watermark does not prove human authorship.
