OpenAI to Implement EU Text Watermarking

OpenAI will introduce invisible watermarking for text and code generated by ChatGPT and Codex in the European Union over the coming weeks. The system, called textGrain, embeds statistical signals into word choices to comply with the EU AI Act. This feature will be mandatory for eligible users in the region, while API customers globally can opt in starting October 5, 2026.

Access to the detection tool is currently limited to approved researchers and expert organizations. The company stated that the watermark is not a global default at launch. Detection efficacy decreases significantly with edited text or short passages, with success rates dropping from 95 percent for 400-token segments to 17 percent when 25 percent of words are replaced.