Stories
OpenAI Introduces Hidden Watermarks for ChatGPT Text in the European Union
N
News18•06-10-2026, 10:16
OpenAI Introduces Hidden Watermarks for ChatGPT Text in the European Union
- •OpenAI is rolling out an invisible watermark called textGrain for ChatGPT and Codex text in the European Union to comply with the EU AI Act.
- •The watermark is hidden from readers but can be recognized by OpenAI detection tools, achieving high accuracy rates under ideal conditions.
- •The system is not foolproof as editing, substituting synonyms, or using short passages can weaken or bypass the watermark detection.
- •Access to the detection tool is initially restricted to approved researchers on a case-by-case basis due to potential false positives and negatives.
- •Developers globally using OpenAI API can opt to turn on watermarking, while other tech giants like Google and Anthropic implement similar measures.
Key Questions Answered
OpenAI is implementing invisible watermarks, called textGrain, to comply with the European Union's AI Act, which mandates that AI-generated text must include a machine-readable marker identifying its origin.
The watermark is an invisible pattern built into the text that OpenAI's detection tools recognize. Under ideal conditions, the detector can identify about 80% of watermarked text at roughly 200 tokens and about 95% at around 400 tokens.
No, text generated by ChatGPT in India will not automatically carry the watermark for now, as there are no similar requirements in India or the US yet.
