OpenAI is adding text watermarking in ChatGPT and Codex
The Verge Stevie Bonifield ● Covered by 5 sources
OpenAI is adding invisible text watermarks to ChatGPT and Codex, starting in the EU. It’s a quiet move, but it shows the AI Act is already shaping how these tools behave.
Based on reporting by The Verge, Stevie Bonifield — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
OpenAI is rolling out an invisible, machine-readable watermark in text from ChatGPT and Codex, with the first users getting it in the European Union. This is not a flashy launch. It is the kind of change most people will never see, which is exactly the point.
The company says its textGrain system “matched or exceeded” other text watermarking approaches, including Google DeepMind’s SynthID for text. That matters because SynthID is also the basis for the watermarking Anthropic announced in August. In other words, the watermark race is already turning into a small standards battle, and the EU is the forcing function.
OpenAI tied the rollout to the EU’s AI Act, and Anthropic made the same basic move for the same reason. So the watermark is not just about technical elegance or product polish. It’s about compliance, and about showing regulators that AI-generated text can be tagged in a way machines can read even if people can’t.
The company also published benchmark scores suggesting watermarked text performs about as well as unwatermarked text. That is the reassuring part. The less reassuring part is what OpenAI itself concedes: textGrain does not guarantee the problem goes away. Watermarks are a tool, not a lock, and the gap between the two is where most of the real debate lives.
My take — AI-written commentary, not fact-checked reporting
This is the AI industry’s favorite trick: call a compliance feature a safety breakthrough and hope everyone applauds. Watermarking text is useful, but it is still a patch on a system that can already mint endless believable words. Europe is doing what it usually does here — dragging the future toward paperwork, one invisible tag at a time.
Read more about this at: The Verge
