Supporting Europe's Efforts to Build a Trustworthy AI Ecosystem

OpenAI announced that it supports the Code of Practice on Transparency of AI-Generated Content published by the European Commission. This is an important step toward implementing the EU AI Act and building a more transparent digital ecosystem.

OpenAI has been working on multiple fronts to strengthen Provenance technology that clarifies the origin of content. Starting with the introduction of C2PA metadata in DALL·E 3 in 2024, OpenAI has laid the technical groundwork through improved marking and detection methods and the release of public verification tools.

Provenance signals indicating the origin of content provide information about the content's creation and editing process, helping users trust the information. This also helps prevent the spread of misinformation and protects the integrity of elections.

To implement stronger provenance, OpenAI has adopted the following multi-layered approach.

• C2PA Content Credentials: Applying C2PA metadata to images generated and edited with DALL·E 3
• Use of multiple signals: Using SynthID watermarking alongside metadata to strengthen signal durability
• Public verification tool: Operating openai.com/verify, allowing users to directly check the provenance signals of images
• Contribution to open standards: Leading the development of interoperable standards through participation in the C2PA Steering Committee