Today for AI
HOT RADAR
industryHEAT 9.6°

AI CLUSTERED EVENT · 10/5/2026

OpenAI Adds Invisible Text Watermark in EU for AI Act Compliance

3 reports archived2 independent sourcesupdated 10/6/2026, 8:50:32 PM
Synthesis & Latest Updates
2 Sources Cross-Validated

OpenAI announced it will add invisible watermarks to text generated by ChatGPT and Codex in the EU to comply with the EU AI Act's transparency rules. The watermark works by subtly shaping word choices rather than using visible symbols, travels with copied text, and is currently default-on only for EU users while global API developers must opt-in.

LATEST/OpenAI announced it will add invisible watermarks to text generated by ChatGPT and Codex in the EU to comply with the EU AI Act's transparency rules. The watermark works by subtly shaping word choices rather than using visible symbols, travels with copied text, and is currently default-on only for EU users while global API developers must opt-in.

TIMELINECoverage timeline

Total 3 reports · Latest first
  1. Ars Technica AIT2·68 pts

    OpenAI Enables Default Text Watermarking for ChatGPT in EU

    Original: OpenAI will watermark ChatGPT outputs by default—but only in the EU

    • OpenAI introduces 'textGrain', a proprietary text watermarking technique enabled by default in the EU due to regulatory requirements.
    • The method works by subtly altering token probability distributions to embed statistical patterns invisible to humans but detectable by machines.
    • Both existing standards (SynthID, C2PA) and OpenAI's solution remain vulnerable to simple circumvention techniques.
    • Access to the detection tool is gated, available only to approved researchers and organizations via a request process.
  2. TechCrunch AIT2·68 pts

    OpenAI Adds Invisible Text Watermark in EU for AI Act Compliance

    Original: OpenAI will start watermarking ChatGPT’s text in the EU

    • OpenAI uses an invisible watermark based on word choice patterns rather than visible markers, ensuring persistence through copy-pasting.
    • Implementation is geographically restricted: default-on for EU ChatGPT/Codex users, but requires manual opt-in for global API developers.
    • The company claims no meaningful performance degradation and confirms the watermark does not identify individual users.
  3. OpenAI 官方博客T1·42 pts

    OpenAI Details EU Text Watermarking Compliance Plan

    Original: Our approach to EU text provenance rules

    • OpenAI launches a phased text watermarking plan in response to the EU AI Act
    • Global API users can now voluntarily enable text watermarking for select models
    • Invisible watermarks will be default for ChatGPT and Codex outputs in the EU within weeks