Security · Planned
ChatGPT plans invisible text watermarks for EU outputs
OpenAI plans to add an invisible statistical watermark to eligible text generated by ChatGPT and Codex in the European Union.
Rollout & details
Announcement date: October 5. Rollout is planned over the coming weeks for eligible users across all ChatGPT plans in the European Union; it is not a global default at launch. The textGrain signal changes word-choice probabilities and is not visible. Detection is imperfect and does not identify the user, measure human contribution, establish ownership or responsibility, or verify accuracy. Detector access begins with approved researchers and expert organizations.
VERIFICATION
Go to the original source.
This report summarizes OpenAI’s published information. Availability may change after this entry’s date.
Read OpenAI’s source