OpenAI will start watermarking ChatGPT’s text in the EU

What happened: OpenAI will add an invisible watermark to ChatGPT and Codex text in the EU to comply with the EU AI Act. The watermark rolls out over coming weeks in the EU only; API access is opt-in. Key numbers: EU AI Act transparency rules took effect August 2. Replacing 10% of words with synonyms dropped detection from about 92% to 66%. Why it matters: OpenAI is not making text watermarking a global default at launch, and says short, edited, or translated text is harder to detect. Anthropic applied its Claude watermark worldwide.

OpenAI will add an invisible watermark to ChatGPT and Codex text in the EU to comply with the EU AI Act. The watermark rolls out over coming weeks in the EU only; API access is opt-in.

OpenAI is not making text watermarking a global default at launch, and says short, edited, or translated text is harder to detect. Anthropic applied its Claude watermark worldwide.