OpenAI announced it will begin adding invisible watermarks to text generated by ChatGPT and Codex in the European Union, a move to comply with the EU AI Act's transparency rules that took effect on August 2.
The watermark is not a visible symbol but works by subtly shaping the model's word choices, leaving a statistical pattern that readers cannot see but a detector can identify. Because it lives in the words themselves, it travels with the text when copied and pasted. OpenAI says the watermark does not identify the user, and that it observed no meaningful change in model performance with it enabled.
The company published a technical report for its method, called textGrain, co-written with researchers from the University of Pennsylvania and Yale. The approach uses a secret key to sort next-word predictions, making hundreds of tiny nudges that together create a detectable signature. The watermark will roll out to EU users over the coming weeks; API developers worldwide can enable it today for select models.
OpenAI acknowledged the watermark's limitations. Tests show that replacing just 10% of words with synonyms drops detection from about 92% to 66%. Short passages, math answers, and translated text are harder to detect. The company said a missing watermark "does not prove human authorship." Anthropic began watermarking Claude text globally two months ago, drawing backlash from users who argued their own instructions and context deserved credit.
