OpenAI Rolls Out Text Watermarking for ChatGPT and Codex Across EU in Phased Implementation

OpenAI has announced a phased rollout of text watermarking for ChatGPT and Codex across the European Union, starting in the coming weeks, as the company moves to comply with the EU AI Act's transparency requirements for AI-generated content. The invisible watermarking system, called textGrain, embeds a statistical signal into the words selected by the model, allowing a detection tool to identify whether content contains an OpenAI watermark. The feature will launch for all ChatGPT subscription tiers and Codex in the EU, while API customers globally can opt in to watermarking on select models—though the setting remains off by default.