European Unionorganization
DevelopingResearch

OpenAI Starts Embedding Invisible Watermarks in ChatGPT and Codex Outputs for European Users

OpenAI has begun inserting invisible watermarks into ChatGPT and Codex outputs for European users and built a detection tool to verify AI-generated text.

Why it matters

The deployment represents a concrete step toward technical AI content identification in a major regulatory jurisdiction. Its significance depends on adoption breadth and whether the detection tool proves reliable against evasion attempts.

Key facts

OpenAI has begun inserting invisible watermarks into ChatGPT and Codex outputs for European users

Reported by Bitcoin Insider

The watermarks are embedded directly into generated output but are not visible to readers

Reported by Bitcoin Insider

OpenAI has built a detection tool to verify when text was produced by its models

Reported by Bitcoin Insider

The watermarking system works by altering the underlying structure of generated text in imperceptible ways

Reported by Bitcoin Insider

The marker is designed to persist within the text and be checked later using OpenAI's detection tool

Reported by Bitcoin Insider

What happened

Bitcoin Insider reports that OpenAI has started embedding invisible watermarks into text generated by ChatGPT and its coding assistant Codex for users in Europe. The watermarks alter the underlying structure of generated text in ways imperceptible during normal reading and are designed to persist within the text itself. OpenAI has also prepared a companion detection tool to flag whether content originated from its models. The move comes after years of debate over distinguishing human-written from AI-generated content, with concerns around misinformation and academic dishonesty. Europe has led regulatory efforts on AI transparency, including the EU AI Act's disclosure requirements.

How the story developed

  1. First report

    First report by Bitcoin Insider

    OpenAI Begins Embedding Invisible Watermarks in ChatGPT and Codex Outputs Across Europe

    Bitcoin Insider

  2. Update

    OpenAI begins watermarking ChatGPT and Codex outputs in Europe

    Invisible watermarks deployed for European users alongside detection tool

    Bitcoin Insider

Affected entities

Original sources

PublisherReportRolePublished
Bitcoin Insider
Aggregator
OpenAI Begins Embedding Invisible Watermarks in ChatGPT and Codex Outputs Across EuropeOriginal