10 Questions for the Content Authenticity Initiative announced at Adobe MAX by Adobe, the NYT, and Twitter

Yesterday at the annual Adobe MAX, the New York Times, Adobe, and Twitter — together, they cover the gamut from creation to distribution — announced the Content Authenticity Initiative. CAI aims to develop an “industry standard for digital content attribution.”

This sounds exciting.

We, at Amber Video, have been vocal advocates these past few years on the need for audio and video provenance and in a world where fakes appear real and carry consequences. (We actually presented two weeks ago at the notable video engineer conference Demuxed on how the video community needs to come together on authentication frameworks and standards.) This specifically means we need to fingerprint content, especially that which has evidentiary value, at source and track it through to distribution, making sure it has not been manipulated along the way. When viewed, the audience should be able to easily see if any segment of the audio/video has been altered.

We hope this is where CAI goes.

From the press release it sounds like CAI is in its infancy with a lot to work through but nevertheless, there are numerous questions that we should preempt now.

Here are ten questions for CAI:

1. When does attribution begin? What if the source can’t be verified?
2. What will attribution entail? Every change made to, say, a Photoshopped image or a video in post production? Can a party follow the steps back to the origin content?
3. Will attribution be tied to a user’s real identity or their device(s)? If not, is attribution worth anything?
4. Can a user remove attribution post-facto? What are the consequences for and against this?
5. Is attribution destructible if the image/video is, for example, subsequently deepfaked?
6. Will the attribution database be centrally managed? If so, by who? Who has write access? Who has lookup access?
7. How will Twitter use attribution in its distribution algorithms?
8. Will Twitter allow its users to look up a post’s attribution?
9. Will Twitter downrank, or even remove altogether, content that doesn’t have attribution?
10. The NYT’s quote suggests this a way to combat “fake news” — can text, and the claims they make, have attribution and even if so, how will this combat “fake news”?

There are many, many more questions for the Content Authenticity Initiative. We look forward to hearing more and working through them together to prevent a world where people can easily dismiss content they don’t agree with as fake just because veracity can’t be unequivocally shown.