OpenAI Implements Invisible Watermark in ChatGPT (Europe)

OpenAI has introduced a new technique for adding an invisible watermark to ChatGPT and Codex texts within the European Union, in compliance with the European AI Act requirements. The watermark relies on the "textGrain" technology, which embeds an invisible statistical signal within the text, making it easier to automatically identify without significantly impacting the model's performance. Tests have shown that this watermark does not greatly affect the quality of the output; however, its detection capability diminishes when texts are modified or edited, especially as the length of the text decreases. The company warns that the watermark is not an infallible method for determining whether a text was generated by AI or a human. Currently, the tool is used solely for research and development purposes, with plans for gradual deployment in Europe.

articleDetails.fontSize.sizeحجم النص

تنويه: هذا ملخص تم إنشاؤه بواسطة الذكاء الاصطناعي

عرض المقال بالكامل