🇳🇱 Dutch, 6 October 2026, AI watermark
ChatGPT Texts to Get Invisible Watermark
ChatGPT texts in EU to be embedded with invisible watermark to detect AI-generated content
The watermark, called textGrain, will be integrated into AI texts from ChatGPT users in the European Union, making it difficult for normal users to distinguish between authentic and AI-generated content. The watermark is not a physical stamp or special symbol, but is embedded in the word choices of the generative AI model. A specialized detector can pick up the watermark, allowing AI-generated words and texts to be identified.
OpenAI, the company behind ChatGPT, says the system is strong but not foolproof, and that short and edited texts can make detection challenging. The company is following the new EU AI transparency regulations, which require AI-generated content to be marked with a watermark.
Anthropic, the company behind chatbot Claude, has also implemented a similar watermark, which is invisible to humans but detectable by computers. The move is seen as a response to the growing competition in the AI industry, with some experts suggesting that companies are adding watermarks to avoid explaining why they didn’t do so earlier.