What are Content Credentials? Here's why Adobe's new AI keeps this metadata front and center
Adobe has been at the forefront of the generative AI space, unveiling many AI solutions, from your classic text-to-image generator to tools geared specifically for enterprises. Regardless of the application, one thing remains the same: the presence of Content Credentials.
At Adobe Summit this week, the company expanded Content Credentials into its Creative and Enterprise applications, including its new Structure Reference in Firefly and Adobe Experience Manager.
Also: This new AI tool from Adobe makes generating the images you need even simpler
This addition highlights Adobe's continued commitment to helping audiences differentiate between authentic and fake content -- an issue exacerbated by the growing presence of generative AI tools -- by including tamper-evident metadata that includes information about how the content was created.
"The technology problem here is the easiest to solve -- we've solved it; adoption is the next hardest thing to solve, but we're showing traction," said Andy Parsons, senior director, Content Authenticity Initiative at Adobe, to ZDNET.
At the Summit, Adobe announced that PricewaterhouseCoopers (PwC), NBCUniversal News Group, WPP, and Autodesk have joined the Content Authenticity Initiative, an organization that promotes the "adoption of an open industry standard for content authenticity and provenance." During the past year, tech giants such as OpenAI, Meta, and Google have supported the use of Content Credentials.
Also: Want to work in AI? How to pivot your career in 5 steps
"You see these large brands coming on board not to just give us their logo and say this is a good idea, but actually implement Content Credentials," added Parsons.
What are Content Credentials?
So, why are these companies all in on Content Credentials? Let's look at Content Credentials and how you can use them.
Content Credentials serve as a "nutrition label" for digital content, allowing users to see the content's vital ingredients, including the creator's name, time and date created, tools used to create the content (such as camera or software program), and edits made.
The information is permanently added to the content through cryptographic metadata and watermarking, ensuring the credentials remain attached, even if someone changes the data or content.
Also: The best AI image generators of 2024: Tested and reviewed
Since Content Credentials are based on free, open-sourced technology, creators can choose to attach Content Credentials to any content to ensure that proper credit is attributed. Additionally, many AI applications -- such as Adobe Firefly -- automatically implement Content Credentials to generate content.
Consequently, it is significant that major players in the generative AI space, such as OpenAI and Meta, have shown a commitment to using Content Credentials to label images generated in their AI models, working towards a world in which all AI-generated images are labeled properly.
Why labeling AI-generated images is so important
As generative AI tools become more popular, and creating realistic-looking fake images becomes easier than ever, the need to properly identify AI-generated content grows more critical.
"As the [generative AI] tools become democratized, and low cost or free, and the quality of those low-cost or free tools produce becomes indistinguishable from reality, of course, [Content Credentials are] evermore important," said Parsons.
Also: We're not ready for the impact of generative AI on elections
Malicious actors can use generative AI to create fake content to spread misinformation and mislead the public. This trend poses a particularly imminent threat in an election year.
"A real concern [is] that perhaps these elections this year in 2024 will be the season of AI elections, where people's understanding of what's real, what candidates are doing, what they're not doing ... [there] will be enough deceptive content [that will] add a lot of noise to the system," added Parsons.
To tackle that threat, and in addition to Content Credentials, Adobe Firefly takes other measures to ensure that AI generations of political figures, or celebrities, aren't possible.
Also: 6 AI tools to supercharge your work and everyday life
Adobe AI was only trained on Adobe Stock Images and public domain content with expired copyrights. The technology can't generate an image of a celebrity or political candidate because it hasn't been trained on this data.
"So between those checks, the ability to add that metadata in, and the way that the image model is trained on openly licensed content -- we're not scraping the open internet -- the Firefly image model couldn't create a Mickey Mouse or a Donald Trump because it's never essentially seeing a Mickey Mouse or Donald Trump," said Deepa Subramaniam, Adobe's VP of product marketing to ZDNET.
Even for applications that don't prohibit the generation of public figures and that are trained by scraping the internet, the presence of Content Credentials can help tackle the issue of provenance and content authenticity by making it easy for users to verify the source of an image, without having to know anything about metadata.
How you can use Content Credentials
For the consumer, getting started using Content Credentials is easy. If you see the Content Credentials pin next to a piece of content, you can click on it to be presented with all of the image's most vital information, as seen in the image at the top of the article. When a pin is unavailable, you can upload the image to the Content Credentials website.
Once the photo is uploaded, if there are Content Credentials within the metadata, they will be shown to you. If not, it will match your image to similar images on the internet and let you know whether or not those images were AI-generated.
Also: Adobe announces generative AI tools to reinvent ad campaigns
I tried the technology by screenshotting the viral image of Pope Francis wearing a puffer jacket and uploading it to the Content Credentials site. Because it was a screenshot, the image didn't have metadata. However, the site did match the photo to other images found on the internet and identified that it was generated by AI, as seen in the image below:
Using Content Credentials was intuitive, easy, and helpful in finding out what I wanted to know, and it will become easier with time as the Content Credentials pin becomes present across the internet.
"I think the best way as a consumer myself to learn about things is to see it in action -- touch it, click it, see how it works," said Parsons. "So, all those resources are available on contentcredentials.org."
Disclosure: The cost of Sabrina Ortiz's travel to Las Vegas for Adobe Summit was covered by Adobe, a common industry practice for long-distance trips. The judgments and opinions of ZDNET's writers and editors are always independent of the companies we cover.