OpenAI has announced that to meet the content transparency requirements of the EU AI Act, it will embed a machine-readable invisible text watermark named textGrain in eligible ChatGPT and Codex text outputs in the EU region within the coming weeks. This technology embeds invisible statistical signals during the model's word selection process, enabling detectors to identify whether the text carries an OpenAI watermark. Initially, this feature will not be set as a global default option, but global API customers can now choose to enable watermarked outputs for select models. OpenAI is collaborating with cloud partners to advance the implementation of the watermarking feature. Meanwhile, researchers and expert organizations can now apply for access to the watermark detector, which, upon approval, can be used to evaluate and optimize text sourcing. The tool only reports watermark detection results and does not disclose user identities, prompts, or conversation content.