OpenAI outlines approach to EU text provenance rules with watermarking

OpenAI has published a statement describing its approach to upcoming EU rules on text provenance, emphasizing the use of watermarking to mark AI-generated text. The company states that watermarks will be applied in specific contexts within its text outputs, aligning with the EU’s focus on traceability for synthetic content.

The announcement highlights how detection of these watermarks will function, outlining mechanisms intended to verify the origin of text flagged as AI-generated. OpenAI frames these measures as part of its response to evolving regulatory expectations in the EU.

Access to the watermarking system is positioned as researcher-first, indicating a phased or controlled rollout that prioritizes evaluation by experts before broader availability. This approach may reflect both technical caution and a strategy to build trust through transparency.