OpenAI is introducing invisible watermarks embedded within ChatGPT-generated text for users in the European Union, according to The Decoder. This move aims to help identify AI-generated content while maintaining user privacy. However, API customers worldwide have the option to disable these watermarks if they choose.
Tests cited by The Decoder reveal that the watermark detection rate can reach up to 95 percent under normal conditions. Yet, this rate drops significantly to 17 percent if approximately 25 percent of the text’s words are altered, indicating some limitations in the watermark’s robustness.
This development is notable for Japanese markets as AI-generated content increasingly influences FX, crypto, and equities analysis, raising questions about transparency and authenticity in automated reports and trading signals.