OpenAI will start watermarking ChatGPT's text in the EU (techcrunch.com)

OpenAI announced that it will introduce an invisible watermark for text generated by ChatGPT and Codex in the European Union, aligning with the EU AI Act's transparency regulations. This initiative, set to roll out over the coming weeks, mandates that AI-generated content be identifiable by AI systems. The watermark subtly alters word choices, embedding an invisible pattern that detects AI-generated text without compromising model performance. Developers worldwide can activate this feature for select models, though it remains off by default. This watermarking technology, dubbed textGrain, was developed in collaboration with researchers from the University of Pennsylvania and Yale, detailing a novel method for identifying AI-generated content using a secret key to guide word prediction. However, it is not foolproof; editing text or using synonyms can reduce detection rates significantly. OpenAI emphasizes that the absence of a watermark does not confirm human authorship, as various factors like length and editing can influence detection. This move by OpenAI comes in the wake of similar actions by competitors and highlights a broader industry commitment to regulatory compliance in AI-generated content.