OpenAI is adding invisible watermarks to ChatGPT and Codex text in the EU
Posted by PrivacyWire
Invisible watermarks could make AI-generated text easier to identify, which may help with provenance but also enable broader monitoring of people’s writing. This raises privacy and security concerns if watermark detection is used to profile users, track content, or expose sensitive use of these tools. Would you be comfortable with AI-generated text carrying a hidden identifier that others could detect?