OpenAI Puts Invisible Watermarks in AI Text for EU Rules

OpenAI is now putting invisible watermarks inside text created by its AI tools, like ChatGPT. This is happening first in the European Union because of new laws there. The watermark is like a secret code hidden in the words. You won't see it, but a computer program can detect it. So if you ask ChatGPT to write an email, essay, or social media post, that text might carry a hidden label saying it was made by AI. The idea is to help people know if something was written by a machine, which could fight fake news and cheating.

But here's the catch: only researchers can access the tool to check for these watermarks. Regular people can't scan a document to see if it's AI-written. Also, watermarks aren't perfect. Someone could edit the text, run it through another AI, or copy-paste it in a way that removes the mark. So it's not a magic solution. OpenAI says this is just one step. They're working on making it more reliable and eventually might let more people use the detection tool.

For you, this means that if you use AI to write something, it might have a hidden label. That could matter if you're a student, a journalist, or someone who writes for work. It might also help you trust what you read online a bit more, but only if the detection becomes widely available. Right now, it's mostly for researchers and regulators. The EU is pushing for this kind of transparency, and other countries might follow. So this could become a global standard for labeling AI-generated text. It's a small step toward keeping AI honest, but it's not a complete fix.