Adobe's New App Protects Your Creations From AI, and I Hope It Sticks

Yahoo

24-04-2025

PCMag editors select and review products independently. If you buy through affiliate links, we may earn commissions, which help support our testing.
At its London Max conference, Adobe announced several new AI models, including Firefly Image Model 4, Firefly Image Model 4 Ultra, and Video Model, all of which promise to make AI-generated content more realistic than ever. The company also said that its products will now be able to use generative AI models from other big players in the space, including Google and OpenAI.
But one announcement at the event stood out to me, as its intention is not to foster unbridled AI content generation, but rather to rein it in. The company introduced the free , which lets creators protect their work by adding secure attribution metadata. It verifies the creators' identity and social media accounts and adds a signal as to whether AI models should use the work for training purposes. The app works by attaching metadata described by the Content Authenticity Initiative, which Adobe spearheaded but enjoys support from members like The Associated Press, Canon, Shutterstock, and The Wall Street Journal.
I fully believe that this initiative is just as important as all the new AI models that can create ever-more-realistic images and video, and I don't see much about it from other AI companies. It reminds me of Mozilla's Do Not Track initiative for browsers from a decade ago, which was meant to protect user privacy. I desperately hope this content authenticity drive succeeds where Do Not Track failed, as it's the only way artists and digital creators can thrive in the age of AI.
Adobe is smart to get involved with this issue. After all, its customers, the artists and designers who use Photoshop, Illustrator, and Premiere Pro, are among those most worried about AI stealing their work. At the London Max event, Adobe's VP of product marketing for Creative Cloud Deepa Subramaniam said, 'One of the biggest concerns we hear from the creative community today is the misuse and representation of their work online.' She positioned the Content Authenticity labels as the modern, secure equivalent of the way artists used to sign their work.
When PCMag's senior security analyst Kim Key dug into this issue, she wrote that the Content Credential metadata 'reveal[s] cryptographic data that verifies the validity of an image and confirms when and how the content was created.' Andy Parsons, Adobe's Content Authenticity director, told Key that "math doesn't lie, people do." This makes a lot of sense.
Previously, you could only apply the Content Credential metadata if you had an expensive Leica camera or exclusively used Adobe's Creative Cloud applications. The new app, now in public beta, allows anyone to apply them. The system applies a "pin" or clickable icon to view info about the contents' provenance and any subsequent edits.
There's still one problem with the system: Adoption. The Content Credentials pin that the app applies is by no means universal. PCMag senior reporter Emily Forlini argues that the other AI vendors, including Apple, Google, and Microsoft, need to label as such, and I couldn't agree more.
This was the downfall of the aforementioned Do Not Track initiative and the Electronic Frontier Foundation's . Both went nowhere because big tech didn't embrace them. Google, in particular, just about single-handedly killed Do Not Track by dissuading users from even turning it on in Chrome. Websites didn't respect the Do Not Track signal either, presumably in the interest of ad profits. This all seems awfully familiar with how AI trains on artistic works without permission.
I can only hope that Adobe's Content Authenticity initiative doesn't suffer the same fate. Some rays of hope include the company's involvement with other interested parties. Subramaniam notes that Content Authenticity leverages LinkedIn's Verified on LinkedIn feature to let creators verify their identity in the new app, for instance.
But AI vendors still need to buy in. That includes Google, Meta, Microsoft, OpenAI, and especially like Midjourney and Stable Diffusion. Unfortunately, Subramaniam had nothing to say on that front. If the big AI vendors don't respect Adobe's "do not train" tag, the initiative could be for naught.
Let's hope history doesn't repeat itself, and that artists and designers get the protections they deserve.