With the rise in popularity of AI-generated images and AI-generated modifications of photographs, there has been an increased desire for transparency regarding whether AI was used wholly, or even in part, in creating an image. To aid in making this determination, Adobe is expanding the capabilities of its Content Authenticity Initiative (CAI) by announcing the creation of a web-based tool called Adobe Content Authenticity.
According to the CAI website, Content Credentials allows a creator to “record and display the most important details about a piece of content at every step of its lifecycle.” These details include the name of the creator and information about the devices used in creating the image, including any software or AI-based tools that were used to modify the image.
Adobe likens Content Credentials to a nutrition label for imagery that helps a viewer understand how an image has been modified from its original incarnation and if that original incarnation was, in fact, AI-generated. I spoke with Andy Parsons, Senior Director of the Content Authenticity Initiative at Adobe, about the Content Authenticity web app, which will launch in early 2025 as a free public beta.
A journalist writing an article about a military conflict would need to be certain that an image accompanying their article was not AI-generated. Conversely, a photographer who took a photo of that content would want editors to know that the image is, in fact, an accurate depiction of a real incident.
Images that utilize Content Credentials would display a small white circle with the letters “cr” inside, known as a CR pin. A viewer would click this pin to access the information associated with that image.
Currently, the Leica M11 is the only camera to seamlessly integrate content authentication using the Content Credentials standard, but there are over 3,700 members of the CAI, including Sony, Canon, and Fujifilm. Platforms such as Google, Microsoft, Meta, and OpenAI will also support the use of Content Credentials.
The new product is called Adobe Content Authenticity. It’s a free web-based tool that allows creators to attach Content Credentials to their work. Once the information has been entered into the image through the Adobe Content Authenticity platform, it will travel with the image across the internet on platforms such as Behance, Instagram, LinkedIn, and X.
“Adobe Firefly, which is our Gen AI set of models and tools, is only trained on content that we have rights to, never on customer content. But we know that every company doesn’t behave that way, and in many cases, companies crawl the open web, waiting for copyright legislation and other things to catch up. We think that creators should have a way to express agency in how they want their material to be used. So we have this setting that also gets imbued in that content credential that says, ‘I don’t want my material to be used to train Gen AI models or to make things that look like my material, thereby co-opting my style, for example,’” said Andy.
While creators may have to wait until next year for the full launch of the Adobe Content Authenticity web app, a Chrome extension offering some of the same functionality will launch today in beta form. The extension will allow anyone to see Content Credentials anywhere they exist on the web. Information about the image, including who created the work, who contributed to the creation, how the image was edited in Photoshop, how the photograph was composited, and if the image can be used to train AI, will be instantly available.
Currently, many photographers use IPTC metadata to embed their photo credit into their images. This information can be intentionally stripped using Lightroom and other products. The information is often stripped unintentionally as an image is modified and transmitted. To ensure this does not occur with Content Credentials, an invisible watermark is added to the image. This undetectable watermark remains with the image even after the image has been screenshotted. The Chrome extension can view this watermark and access the information it contains.
Creators who wish to be part of the public beta for the Adobe Content Authenticity web app can sign up here.