AI Transparency Through C2PA
As AI image generation becomes mainstream, major platforms are adopting C2PA to transparently label AI-created content. Here's how each platform implements it:
OpenAI DALL-E
DALL-E embeds C2PA metadata in every generated image. The manifest includes:
- digitalSourceType:
trainedAlgorithmicMedia— indicating the image is fully AI-generated - Claim Generator: Identifies DALL-E as the creation tool
- Timestamp: When the image was generated
Adobe Firefly
Adobe was a founding member of the C2PA coalition. Adobe Firefly embeds comprehensive Content Credentials including:
- Full creation history with AI model information
- Adobe's digital signature
- Integration with Adobe Content Authenticity Initiative (CAI)
Google Gemini
Google's Gemini (formerly Bard) image generation includes:
- C2PA metadata with Google's certificate
digitalSourceTypeidentifying AI generation- SynthID watermarking as an additional layer
Microsoft Designer / Copilot
Microsoft embeds C2PA metadata through its Designer and Copilot image generation tools, certified with Microsoft's signing certificate.
How to Verify
Upload any AI-generated image to C2PA Checker to see the embedded metadata. The tool will extract the manifest, show the digitalSourceType, identify the AI tool, and verify the digital signature.
Midjourney — The Exception
As of early 2026, Midjourney does not embed C2PA metadata in its generated images. This makes Midjourney images harder to identify through provenance checking alone, requiring alternative detection methods like pixel-level forensic analysis.
