OpenAI has published its approach to text provenance requirements under EU rules, focusing on text watermarking. The post describes where watermarks are applied, how detection of watermarked text works, and the decision to begin access with researchers.
Why it matters
Text provenance and watermarking are central to current EU regulatory efforts around AI-generated content. OpenAI’s stated approach signals how a major model provider intends to align with these rules and support verification of AI-produced text.
Who should care
Researchers are the first group to receive access to detection, according to the post. Policymakers and others following EU AI regulation may also be interested in how watermarking and detection are being implemented in practice.