Adobe is asking companies to watermark AI-generated images
The Adobe executive leading an initiative to mark AI-generated images says he hopes Apple and X will soon join the coalition of companies that have agreed to do so.
Content Credentials is an open standard designed to increase the transparency of AI-generated content. It aims to put metadata and put a small “CR” logo in the top corner of images that have been created entirely by AI, making it clear to consumers that an image is not real.
Content credentials are managed by the Coalition for Content Origin and Authenticity (C2PA), whose members include Adobe, Google, Microsoft, OpenAI, Meta, the BBC and many other technology and media companies. However, there are some notable absences from the list of consortium members, including Apple and Elon Musk’s X.
AI images marked in the browser
The success of a scheme such as Content Credentials will depend on widespread adoption, with users being able to see the Content Credentials logo regardless of which device or social network they use.
Andy Parsons, senior director of the Content Authenticity Initiative at Adobe, said it was only a matter of time before support for displaying the CR logo on AI-generated images was built into popular web browsers, such as Google Chrome.
“We’d like to see a more ubiquitous and visible adoption so that consumers can recognize the little icon of Content Credentials and it’s something, I think, that can be just as familiar and widespread than the copyright symbol,” Parsons said at a meeting. – a conference at CES 2025 in Las Vegas last week.
However, cross-device support will depend on Apple agreeing to be part of the scheme, with Safari commanding about a fifth of the global mobile browser market according to Statcounter.
“We hope to see Apple join,” Parsons said. “Apple is conspicuously absent from the consortium, but I hope that will change, perhaps this year.”
“My impression is that Apple tends to be very slow in merging sets of standards,” he added. “Historically, they tend not to do it until something is very well established and they kind of have to.”
Parsons added that he hoped Apple would not create a competing standard for reporting AI content, claiming it would be bad for everyone involved if different watermarks were applied to AI images. “My hope would be that, in the fullness of time, Apple would agree with that view, but we’ll see.”
Apple was approached for comment on its approach to AI imaging.
Musk challenged
Grok AI i X has few limitations in image generation
X is another notable absence from the list of consortium members. The social network has its own Grok AI capable of generating images, but unlike some rival AI services, it doesn’t try to prevent users from generating images of celebrities like Donald Trump engaging in controversial topics , as you can see. from the image above.
When I tried to post one of the images above, it automatically added an “image by Grok” line to the post text, but it was easily deleted. If you download the image, the metadata cites “x.com” as the location where the photo was taken, but this is not immediately accessible to a viewer on the social network.
Twitter was originally a member of C2PA, but “then some things happened and they became less active,” Parsons said.
Parsons said he raised the issue with Elon Musk at an AI conference in London last year, hosted by then UK Prime Minister Rishi Sunak.
“Rishi Sunak called me as the last question before the process ended, and I asked Elon directly, ‘Would it be a good thing for the X platform if there was a way to cryptographically sign the content to be delivered? people an objective understanding of where it came from and who is posting and what it is?’ And he gave this short comical answer. He thought for a second and said ‘that sounds like a good idea, we should do it'”.
Parsons said he remained “hopeful and optimistic” that X would join the consortium, “but they are not members”.
