ChatGPT text gets a hidden watermark in the EU

What happened: OpenAI said Monday that over the coming weeks, eligible ChatGPT and Codex output in the EU will carry TextGrain, a statistical signal hidden in word choices that readers can't see but a detector can. API customers worldwide can turn it on, but it's off by default. OpenAI plans to open source the technology, and at first only approved researchers and organizations get the detector. ChatGPT joins Anthropic's Claude and Google's Gemini in marking its own text. In OpenAI's own test at a 1% false positive rate, the detector found the mark in about 95% of 400 token passages and about 80% of 200 token passages. Swapping just 25% of the words for synonyms dropped detection to 17%.
The claim: OpenAI says this meets the EU AI Act's transparency rules for AI generated content, and that a missing watermark does not prove human authorship.
Austin's take: Austin says this makes little sense for text, since he finds AI writing easy to spot and tools like Pangram already exist. He sees watermarks mattering more for deepfake video, voice and images, and calls this an example of government moving slower than the market.
Why it matters: Teachers, employers and platforms want a way to tell AI writing from human writing, but the numbers show light editing largely defeats the mark.
What to do: Don't treat a missing watermark as proof a human wrote something. If you use the API, decide whether to switch it on.
Not confirmed yet: The accuracy figures come from OpenAI's own testing.
Sources: Full stories and sources

The Loop