OpenAI Will Add an Invisible Watermark to ChatGPT Text in the EU: How It Works

OpenAI will begin adding an invisible watermark to text generated by ChatGPT and Codex for users in the European Union, to comply with the EU AI Act, the company said on Monday. The rollout will happen over the coming weeks for eligible users on all plans in the EU. Developers using OpenAI's API anywhere in the world can also turn the feature on for select models from Monday, although it stays off by default, and OpenAI is not making text watermarking a global default at launch.

How the Watermark Works

The watermark does not appear as a visible symbol. OpenAI's system, named textGrain, uses a secret key to subtly influence the model's word choices, creating a hidden statistical pattern that specialised systems can detect. Because the pattern sits inside the text itself, OpenAI says it should remain detectable after copying and pasting. The company says the watermark does not identify the person who generated the text, and its testing showed no meaningful impact on model performance.

It Can Be Weakened

OpenAI acknowledged the watermark is not impossible to remove or weaken. In one company test, replacing about 10% of the words in a passage with synonyms reduced detection accuracy from about 92% to 66%. Short passages, mathematical answers and translated text can also be harder to identify. For these reasons, OpenAI will initially give access to its detector only to approved researchers and expert organisations.

What a Missing Watermark Means

OpenAI cautioned that the absence of a watermark does not prove a person wrote the content, since a passage may be too short, heavily edited or generated by another AI system. A watermark can show that an OpenAI system generated or processed part of a passage, but it cannot show how much human judgement or creativity was involved.

Why Now

The move follows transparency requirements in the EU AI Act, which require providers of certain AI systems to make AI-generated content identifiable by detection systems. OpenAI is among several major technology companies, including Anthropic, Google, Meta and Microsoft, that have committed to the EU's code of practice on transparency for AI-generated content. Anthropic announced in August that it would introduce watermarking for text from its Claude models globally.