Skip to content

Content Authenticity / Credentials / Provenance tracking issue #652

@jeffpaul

Description

@jeffpaul

Is your enhancement related to a problem? Please describe.

One issue with AI-generated (or augmented) content is that its currently hard for a consumer of that content to know that it was AI generated or not by merely reading/viewing said content. Said more simply, without taking some actions here we're potentially contributing to the erosion of trust in online content and interactions. We have the opportunity with ClassifAI to help (1) track what content is generated or augmented via AI and (2) provide mechanisms to alert users on the front-end when viewing said content. While we cannot prevent deepfakes and similar malevolent actors, it feels like we can still try and help contribute as best we can to improve the trust in content and interactions online.

Designs

As originally referenced in #398 (comment), we can look to provide a way to track what content has been generated or augmented with AI and then display attribution where that data exists. While we have an image generation credit and license in the caption of generated images we could go further with the following:

  • Leverage Content Authenticity Initiative (CAI) tooling to integrate secure provenance signals (aka C2PA data) into image files generated/augment via AI (currently the Node.js option might be the most feasible) noting this could be iterated upon to also include image editing done within the Media Library
  • Displaying C2PA data via a Content Credentials cr overlay on the front-end for images (where said data exists)
  • Expose the C2PA data when viewing images in the Media Library, potentially having that similarly available as a new collapsed-by-default sidebar panel in the image/cover image/etc blocks

While the CAI tooling does not yet appear to support similar tracking and attribution for text-based content, we could still look to add some custom ClassifAI meta to posts where AI tooling is used to track what was done, when, and by which users so that data could be presented or utilize later on (e.g. post footnotes referencing AI features used in the generation of the post content, reporting on what features are most used).

Describe alternatives you've considered

No response

Code of Conduct

  • I agree to follow this project's Code of Conduct

Metadata

Metadata

Assignees

No one assigned

    Labels

    help wantedExtra attention is neededneeds:engineeringThis requires engineering to resolve.

    Projects

    Status

    To Do

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions