ChatGPT to introduce invisible watermarks on generated text

OpenAI has announced that text generated by ChatGPT will soon carry an invisible watermark, starting in Europe. This measure aims to effectively identify content created by artificial intelligence, something that has been complicated in the text domain compared to images or audio, which already have dedicated detection tools.

The implementation of this technology means that while texts generated by ChatGPT may seem human-like, they can be more clearly distinguished through the use of specific detectors. However, it is important to note that edits to the text or the use of shorter texts could make the detection of this watermark more challenging, raising questions about its effectiveness in all cases.

## Why it matters The introduction of watermarks on AI-generated text is a significant step toward regulation and transparency in the use of these tools. This could influence how businesses and individuals utilize ChatGPT, especially in contexts where content authenticity is crucial, such as in academic or journalistic fields.

## What we know We know that OpenAI is working on this watermark technology for text generated by ChatGPT and Codex, and that its implementation will begin in Europe. This represents an effort by OpenAI to address concerns about the misuse of AI-generated content and the need to identify its origin.

## What remains unclear It is still unclear how this watermark will be implemented across different platforms or if there will be variations in its effectiveness depending on the type of text. Additionally, details have yet to be provided on how users will be able to verify if their text carries the watermark.