OpenAI adds invisible text watermarks for EU ChatGPT and Codex users

OpenAI announced on October 5 that it will add an invisible text watermark to content generated by ChatGPT and Codex in the European Union, complying with the transparency rules of the EU AI Act that took effect on August 2. The rollout will begin in the coming weeks for ChatGPT and Codex users on all plans in the EU.

Developers using OpenAI's API worldwide can enable the feature for select models starting immediately, with the setting off by default. The company said it is not making text watermarking a global default at launch. The watermark is not a visible symbol; it works by subtly shaping the model's word choices, creating a pattern that readers cannot see but a detector can recognize. The pattern travels with the text when it is copied and pasted. OpenAI said the watermark does not identify users, and that model performance showed no meaningful change with it enabled.

OpenAI also published a technical report, called textGrain, co-written with researchers from the University of Pennsylvania and Yale. The company's tests found that editing can reduce detection effectiveness: replacing 10% of words with synonyms dropped the detection rate from about 92% to 66%. Short passages, math answers, and translated text are harder to detect. As a result, initial access to the detector will be limited to approved researchers and expert organizations. OpenAI also cautioned that a missing watermark does not prove human authorship.

Anthropic announced in August that it would watermark text generated by Claude, applying the measure worldwide. OpenAI had previously built a text watermark but held off on releasing it, partly over concerns that users would switch to rivals that did not watermark.