(Credit: Adobe)
Adobe has announced a new web app, dubbed Adobe Content Authenticity, that image creators and artists can use to label their work with the goal of retaining artists' authority online. It's also launching a companion Chrome extension that lets internet users check any webpage for existing labels on content.
Adobe Content Authenticity lets users add Content Credentials to their images and videos. Content Credentials is a feature developed by the nonprofit group Coalition for Content Provenance and Authenticity (C2PA). The group has corporate members like Adobe, Google, and Meta—and is trying to get internet users to voluntarily label their AI-generated creations as such.
Adobe's new platform for adding Content Credentials allows creators to add a range of different tags to their work regardless of whether it was AI-generated. The labels can then appear via an expandable button in the top corner of the work on supported platforms. Creators can add their names, websites, social media profiles, and other details, like when the piece of media was created and how exactly it was made or edited.
They can also use Adobe's tool to "signal" if they don't want AI models to train on their work. "Adobe is actively working to drive industry-wide adoption of this preference so that other generative AI models that support it do not train on or utilize creators’ work," the company adds. But, as Adobe says, this is just a "preference," not a guarantee.
Data-poisoning tools like Nightshade, which can garble some AI outputs, may be a more effective solution for concerned artists. Nightshade doesn't ask politely—it effectively forces AI models not to swipe the work by discreetly "poisoning" the art behind the scenes.
As for Adobe itself, the company previously faced controversy over a policy change that allowed it to "access" and "view" users' content on its apps like Photoshop or Premiere Pro—and run that content through its own AI tools. But Adobe said the situation was a big misunderstanding, and it never uses customer-generated content to train its AI models or tools like Firefly. Adobe has also taken action against Adobe Stock accounts that tried to sell AI-generated imitations of famous artists' work.
While Content Credentials is a step in the right direction, it's unclear to what extent the general public will demand credentials on every image they see online. There's also the chance bad actors could misuse the feature to assign an incorrect name or false labels to a piece of work.
(Credit: Adobe)The opt-in feature isn't compatible with every site or social media platform, so Adobe is launching a Content Authenticity extension for the Google Chrome browser. That will let you see whether a piece of art has been assigned a label at some point, but it means you'll have to install the extension (if you use Chrome).
Adobe claims it's difficult to remove Content Credentials from images or videos once they've been added—even if someone takes a screenshot. "Content credentials combine digital fingerprinting, invisible watermarking and cryptographically signed metadata, helping to ensure that Content Credentials remain intact and verifiable across the digital ecosystem," Adobe's announcement reads.
The tool doesn't directly address problem of deepfakes, which are AI-generated images crafted and published with the intent of being purposely misleading. But, if the data in Content Credentials were verified by a trusted third party and became the expectation and the norm across the entirety of the internet, then unlabeled content might become inherently distrusted. That could take some time and would require the cooperation of some of the biggest companies in the world, however.
Adobe's Content Authenticity web app is expected to go live early next year in a free public beta. The Chrome extension is available in beta now.