< Back to all clusters
[TECHNOLOGY] · EU, United States · 100 sources

started · updated

OpenAI introduces textGrain invisible watermarking for EU compliance

OpenAI is introducing textGrain, an invisible digital watermarking technology designed to identify AI-generated content. This move is intended to help the company comply with the transparency requirements of the European Union's AI Act.

In the coming weeks, OpenAI will begin automatically applying these watermarks to text outputs from ChatGPT and Codex for eligible users within the European Union across all subscription plans. Outside of the EU, the technology will be available to API customers worldwide as an opt-in feature, though it will remain disabled by default.

Unlike visible markers, textGrain works by subtly altering the statistical patterns of word selection during the generation process. This method ensures the watermark remains embedded even if the text is copied and pasted. However, the technology has limitations; detection accuracy can be significantly reduced if the text is edited or if words are replaced with synonyms. Currently, access to the detection tools is restricted to approved researchers and expert organizations.

Entities

Anthropic · ChatGPT · Codex · EU AI Act · European Union · European Union · OpenAI · University of Pennsylvania · textGrain

Claims

What the coverage asserts, and how many sources carry each claim.

Sources