OpenAI Will Secretly Watermark ChatGPT Text in Europe
This could help detect AI-written content, but it's not foolproof.
OpenAI has announced that it will automatically watermark text generated by ChatGPT in the European Union. This is to comply with the EU AI Act, which requires AI-generated content to be detectable.
The watermark, called textGrain, works by inserting subtle patterns into the word choices that are invisible to human readers but can be detected by a special tool. However, this tool is not perfect: OpenAI's tests show a 92% detection rate, but if someone edits just 10% of the text, detection drops by 30%, and editing 20% drops it by 75%.
Also, shorter or translated texts are harder to detect. The watermark will be on by default in the EU in the coming weeks, but outside the EU, it will be off by default, though users can opt in. OpenAI's competitor Anthropic already has watermarking globally. This move is part of a broader effort to regulate AI and ensure transparency.
- OpenAI will automatically watermark ChatGPT text in the EU to comply with new AI laws.
- The watermark is a hidden pattern that can be detected by a special tool, but it's easy to break by editing the text.
- Outside the EU, watermarking is optional, so most users won't see it unless they turn it on.
Why It Matters
This could help detect AI-written content, but it's not foolproof and only applies in the EU for now.