OpenAI is rolling out a multi-layered approach to content provenance, addressing growing concerns about AI-generated media authenticity. The company announced C2PA conformance for its tools, making provenance signals easier for other platforms to recognize and preserve across the content lifecycle.

Through a partnership with Google, OpenAI is integrating SynthID watermarking into images created via ChatGPT, Codex, and the OpenAI API. This invisible watermark complements C2PA metadata, providing durability when files undergo transformations that might strip traditional metadata. The dual-layer system addresses limitations where metadata alone can be lost through screenshots, resizing, or format conversions.

OpenAI also previewed a public verification tool that checks uploaded images for both Content Credentials and SynthID watermarks to determine if they originated from OpenAI services. The tool takes a cautious approach, avoiding definitive conclusions when no provenance signals are detected, since these markers can be stripped through various modifications.

This initiative builds on OpenAI's earlier work adding Content Credentials to DALL·E 3, ImageGen, and Sora, plus joining C2PA's Steering Committee. The company emphasizes that no single provenance technique suffices independently, advocating for combined standards, watermarking, and public verification to create a more trustworthy AI ecosystem.