OpenAI, the company behind ChatGPT, has outlined its approach to complying with new European Union (EU) regulations on text provenance. These rules require certain text generated by artificial intelligence (AI) to carry digital watermarks that help identify its source. OpenAI has explained where these watermarks apply, how they can be detected, and why access to detection tools will initially be given to researchers.

This development is significant because the EU’s new text provenance rules aim to increase transparency around AI-generated content. By embedding watermarks, it becomes easier to trace the origin of text, which helps combat misinformation and supports responsible AI use. For readers around the world, these rules demonstrate how governments are beginning to regulate AI-generated information at a technical level.

The new EU rules require companies like OpenAI to mark text created by their AI systems so that it can be identified as generated rather than human-written. OpenAI’s approach involves applying these watermarks directly to outputs from their models. Detection of watermarked text is currently intended to be available mainly to invited researchers, allowing for controlled study of AI text and its impact. This step helps ensure that watermarking technology is reliable before wider public use.

Moving forward, it will be important to watch how OpenAI and other AI providers implement watermarking across different languages and contexts. Wider access to detection tools and further EU guidance could shape how text provenance rules influence AI use globally.