OpenAI begins invisible watermarking of ChatGPT and Codex text in the EU

OpenAI is beginning a phased response to European Union rules that require AI-generated text to be recognizable by software. Over the coming weeks, eligible ChatGPT and Codex output in the EU will receive an invisible statistical pattern created by subtle choices among possible words; a specialized detector, not a reader, can look for it. The aim is to give publishers, platforms, and researchers a clue about where text came from without putting a visible label on every answer—but it is not proof of authorship: short passages, rewriting, or translation can make the signal hard to find. Developers worldwide can opt into the feature for selected API models, while access to the text detector initially remains limited to approved researchers and expert organizations.