According to CoinMeta, OpenAI has announced that it has added invisible text watermarks to its AI systems ChatGPT and Codex in Europe to comply with the transparency requirements of the EU's Artificial Intelligence Act. The act came into effect on August 2nd and requires AI companies to label content generated by machines. OpenAI's watermark system is called Textgrain, and it hides the watermarks by making subtle adjustments to the words selected by ChatGPT. This ensures that the watermarks can still be detected even when the text is copied and pasted. The company stated that the watermarks will not contain any information that identifies users, and that when the watermarks are enabled, there has been no significant decline in the performance of GPT-6 and Astra.