Which AI Image Generators Support C2PA? Midjourney, DALL-E 3, Sora & More

The EU AI Act (effective August 2026) requires transparency labeling for AI-generated content. C2PA Content Credentials are the leading technical mechanism for that disclosure. But adoption across AI image generators is uneven: some embedded credentials from day one, others have ignored the standard entirely. Here is the confirmed status for each major platform.

Adobe Firefly — Full Support Since Launch

Adobe Firefly is the gold standard for AI tool C2PA implementation. As a founding member of both the C2PA coalition and the Content Authenticity Initiative, Adobe built Content Credentials into Firefly from its public launch in March 2023.

Every image generated by Firefly — whether through the standalone Firefly web app, Photoshop's Generative Fill, Illustrator's Generative Recolor, or the Firefly API — carries a C2PA manifest that records:

• That the image was AI-generated
• The specific Firefly model used
• The approximate generation timestamp
• Adobe as the signing organization

Adobe also preserves credentials through its editing pipeline: if a Firefly-generated image is edited in Photoshop and saved, the new manifest references the original Firefly credential as a parent, creating a full provenance chain.

OpenAI DALL-E 3 — Supported Since 2023

OpenAI added C2PA Content Credentials to DALL-E 3 in 2023, working in partnership with the Content Authenticity Initiative. Images generated through ChatGPT's image generation feature (which runs on DALL-E 3) and through the OpenAI API carry a C2PA manifest.

The manifest identifies the content as AI-generated by OpenAI and includes a timestamp. OpenAI signs these credentials using a certificate registered with the C2PA Trust List, making them verifiable by any compliant tool including C2PA Viewer.

OpenAI Sora — C2PA on AI-Generated Video

OpenAI announced C2PA support for Sora, its video generation model, consistent with the company's broader commitment to content provenance across its generative AI products. Sora-generated videos carry a C2PA manifest identifying the content as AI-synthesized.

This is notable because video C2PA support is less common than image support — the JUMBF container format used to embed C2PA manifests requires specific handling for video containers (MP4, MOV), and Sora's implementation covers this correctly according to the C2PA v2.2 specification.

Google Imagen — Supported via SynthID + C2PA

Google has embedded C2PA Content Credentials in images generated by Imagen, its text-to-image model. Google also uses SynthID, its proprietary invisible watermarking technology, alongside C2PA — meaning Imagen outputs carry both a readable C2PA manifest and an invisible pixel-level watermark.

The dual approach addresses a known weakness of C2PA: manifest stripping. If someone downloads an Imagen image and re-saves it through a non-C2PA tool, the C2PA manifest may be lost. The SynthID watermark, embedded in pixel data, survives most re-saves and format conversions. The two signals are complementary.

Midjourney — No C2PA Support

Midjourney does not embed C2PA Content Credentials in generated images as of early 2026. Images downloaded from Midjourney via Discord or the web interface carry no C2PA manifest.

Midjourney has not publicly committed to a C2PA implementation timeline. Given the EU AI Act's August 2026 effective date and its transparency requirements for AI-generated content, this is a notable gap for a platform that serves millions of users generating images for commercial use.

Stable Diffusion — Partial Support (Third-Party)

Stable Diffusion is open-source software, not a hosted service, which makes C2PA support a question of which interface is used rather than a single platform decision. Stability AI itself added C2PA support to its hosted Stable Image platform. The open-source model weights and base code do not embed C2PA manifests by default.

Third-party interfaces for Stable Diffusion (ComfyUI, Automatic1111, InvokeAI) do not add C2PA credentials by default, though they could in principle. If you run Stable Diffusion locally or through a third-party host, generated images will not carry C2PA credentials unless that host has specifically implemented them.

What the EU AI Act Means for AI Tool C2PA Adoption

The EU AI Act, effective August 2026, requires that AI-generated content be disclosed as such when it could be mistaken for human-created content. While the Act does not explicitly mandate C2PA, Content Credentials are the leading technical mechanism for satisfying this disclosure requirement.

Platforms that currently lack C2PA support — Midjourney being the most prominent — face regulatory pressure to implement some form of provenance disclosure before August 2026. Whether they choose C2PA specifically or an alternative approach remains to be seen.

How to Check if an AI-Generated Image Has C2PA Credentials

If you receive an image that may have been AI-generated and want to check whether it carries C2PA disclosure:

1. Drop it into C2PA Viewer: The tool will show whether a manifest is present, who signed it, and what the manifest claims — including whether the content is marked as AI-generated. All processing is client-side.
2. Check the signing organization: A valid C2PA manifest from an AI tool will identify the signing organization (Adobe, OpenAI, Google) in the certificate issuer field.
3. No manifest ≠ definitely not AI: The absence of a C2PA manifest does not prove an image is human-made. It only means the generating tool did not embed credentials — or they were stripped. Midjourney and many Stable Diffusion interfaces produce images with no C2PA data at all.

Frequently Asked Questions