OpenAI is taking a meaningful step toward clearer AI provenance by rolling out invisible, machine-readable text watermarking for ChatGPT and Codex users in the European Union.
The new system, called textGrain, is designed to help identify AI-generated text without visibly altering the user experience. OpenAI says its approach matched or exceeded other watermarking methods, including techniques related to Google DeepMind’s SynthID for text.
Why it matters
As AI-generated writing becomes more common, reliable transparency tools can help educators, platforms, businesses, and policymakers better understand when content may have been produced by AI. This rollout also shows how regulation like the EU AI Act can accelerate real-world safety and accountability features.
- Trust: Watermarking can support clearer labeling of AI-generated content.
- Performance: OpenAI says benchmark scores were similar for watermarked and unwatermarked outputs.
- Deployment: The feature is not just a research demo — it is rolling out to real ChatGPT and Codex users in the EU.
While no watermarking method is perfect, OpenAI’s launch is a practical win for AI transparency. It adds another layer of accountability as generative AI tools become part of everyday work, coding, and communication.