OpenAI to begin watermarking AI-generated text
OpenAI watermarks ChatGPT text—but only by default in the EU. Globally, users can disable it.
Why it matters
OpenAI is implementing digital watermarking for AI-generated text to comply with EU AI Act requirements, but the feature is opt-in outside the EU and reports suggest it can be edited out, raising questions about its enforceability.
The key facts
12 to knowWatermark rolling out within weeks of announcement
Enabled by default only in EU; users elsewhere can disable it
Response to EU AI regulations requiring identifiable AI-generated content
Anthropic shipped similar feature in August 2026
TechCrunch reports watermark appears editable, limiting practical utility
Available worldwide but with regional default behavior
Feature rolling out 'within a few weeks'
Enabled by default only in EU; available but optional worldwide
Competitor Anthropic introduced similar watermark in August 2026
Motivated by EU AI regulations introduced two months prior (August 2026)
TechCrunch reports watermark appears editable, raising effectiveness questions
Applies to ChatGPT and Codex
The story so far
Earlier coverage of this storyline
- OpenAI is adding text watermarking in ChatGPT and CodexThe Verge AI
- OpenAI will start watermarking ChatGPT’s text in the EUTechCrunch AI
- This story
Go to the source
Computerworldcomputerworld.com
Publisher excerpt: OpenAI has announced that within a few weeks it will introduce a digital watermark for text generated using the company’s AI tools, ChatGPT and Codex. This move comes in response to the EU’s new AI regulations, introduced two months ago, which stipulate that AI-generated content must be…