OpenAI to watermark ChatGPT text in the EU
What happened
From 6 October 2026, multiple outlets reported that OpenAI will add invisible, machine-readable watermarks to text generated by ChatGPT and Codex under the EU AI Act, using a technique called textGrain. The reports say the first phase covers EU users only, expanding over the coming weeks to eligible users on all EU plans, and is not yet a global default; global API users can opt in, with the feature off by default, unlike Anthropic's globally mandatory watermarking for Claude. On the compliance basis, earlier reports said it was to meet EU AI Act transparency rules effective 2 August, while later reports said the act took effect in August and requires AI-generated content to be marked in a way other tools can detect. OpenAI has published a technical paper explaining the method, and the detector will first be opened to a limited number of researchers and institutions, with others able to join gradually through an application process.
Written by AI from the reports · updated 9 d ago
Timeline
Follow the reports to see the event from each side.
- Ars Technica · AIOpenAI will watermark ChatGPT outputs by default—but only in the EU
Like other solutions, it is not especially reliable, and it's easy to circumvent.
- TechCrunch · AIOpenAI will start watermarking ChatGPT’s text in the EU
OpenAI will watermark ChatGPT and Codex text in the EU to comply with the AI Act. Editing can make the invisible marks harder to detect, it says.
- The Verge · AIOpenAI is adding text watermarking in ChatGPT and Codex
An invisible, machine-readable watermark in text output is rolling out to ChatGPT and Codex, but only for users in the European Union at first. OpenAI says its textGrain watermarking "matched or exceeded" other approaches like Google DeepMind's SynthID for text, which is also the basis for the watermarking Anthropic announced in August.
- The DecoderOpenAI will watermark ChatGPT text in the EU but makes it optional for API users worldwide
OpenAI is adding invisible textGrain watermarks to ChatGPT in the EU, but unlike Anthropic, it will let API customers worldwide opt out. Tests show detection rates as high as 95 percent, dropping to 17 percent when a quarter of the words are replaced.