OpenAI to add automatic watermarks for ChatGPT users in the EU

OpenAI has announced that it will implement automatic watermarking for text generated by ChatGPT within the European Union. This decision follows the implementation of the EU AI Act in August, which mandates that artificial intelligence-produced content must be identifiable by detection tools. While the company plans to make this feature available globally, the watermarking function will remain disabled by default in all regions outside of Europe.

The company is utilizing a proprietary method known as textGrain, which embeds subtle linguistic patterns into word selections. These patterns are invisible to human users and do not degrade the quality of the text, but they allow specialized software to identify the content as machine-generated. Although OpenAI is providing detector access to select researchers, current industry standards remain susceptible to circumvention. Consequently, organizations should rigorously evaluate the capabilities, data usage, and security protocols of any AI tools before integrating them into their professional workflows.