OpenAI starts watermarking ChatGPT text in the EU
OpenAI is rolling out invisible text watermarking for ChatGPT and Codex in the European Union to comply with the EU AI Act's transparency rules. The method, called textGrain, subtly shapes word choices to leave an invisible pattern that detectors can identify. While developers globally can enable the feature via the API, it is not being made a global default. OpenAI notes that edits, translations, and short passages can weaken detection accuracy, highlighting ongoing challenges in provenance tracking.
