OpenAI to start watermarking ChatGPT text in the EU

OpenAI has announced that it will begin watermarking text generated by its ChatGPT and Codex models in the European Union. This measure is being taken to comply with the new EU AI Act, which aims to regulate the use of artificial intelligence and ensure transparency in its application.

The inclusion of watermarks is a strategy to identify AI-generated content. While these marks are invisible to the naked eye, OpenAI warns that editing the texts could make their detection harder. This raises questions about the effectiveness of the watermarks in contexts where users modify or adapt the texts.

Why it matters

The implementation of these watermarks is a significant step in the regulation of AI usage in Europe. As artificial intelligence becomes more integrated into daily life, it is crucial to establish standards that ensure users can identify when they are interacting with machine-generated content.

Additionally, this measure could influence how tech companies handle transparency and accountability in AI usage, setting a precedent for other regions looking to regulate these technologies.

What we know

OpenAI has confirmed that the watermark will be applied to all text generated by ChatGPT and Codex in the EU, meaning that users in this region will be subject to this new regulation. This action is part of a broader effort to comply with emerging regulations in the field of artificial intelligence.

What remains unclear

It is still unclear to what extent the watermarks will be visible or how the detection process will be implemented in edited texts. There is also a lack of information on how this will affect user experience and whether there will be a difference in the treatment of generated content in other regions outside the EU.