OpenAI Rolls Out ChatGPT Text Watermarking in EU, but Rewriting Defeats It

alex_verem · x · 2026-10-06

OpenAI announced it will start watermarking eligible ChatGPT and Codex text output in the EU over the coming weeks to comply with the EU AI Act, extending its image/audio provenance tools. API customers can enable text watermarking on select models worldwide today.

Critics, citing OpenAI's own admission of limitations, note that rewriting the text removes the watermark: OpenAI's internal tests spot only 80% of 200-token passages, and short text often slips through entirely — meaning it mainly catches users too lazy to paste output into another chatbot.

Related event: OpenAI launches textGrain text watermarking, starting in the EU(25 posts)→

Original post →

More from Models

Models channel →