For the sake of transparency, Adobe has developed a creative icon, for Content Credentials, to mark all content produced by generative AI. A good initiative… provided that everyone adopts it!
Image-generating AI fascinates Internet users as much as it worries them. In just seconds, they can produce stunning images in a wide variety of artistic styles in response to a simple text command. Whether it’s a surrealist painting, a portrait drawn with a Bic pen, or even a strikingly realistic photograph, everything is at your fingertips. Among the different artificial intelligences that stand out for the quality of their renderings, we include DALL-E, developed by the creators of ChatGPT, its competitors Midjourney, Bing Image Creator and Microsoft Designer, or even Adobe Firefly. Even Google got in on the act with Bard! But the results obtained can be so realistic that it is sometimes impossible to distinguish an artificial creation from a 100% original production, or even from reality, which poses considerable challenges in terms of fact-checking (verification of authenticity, in French). It is therefore more than necessary to have tools to determine whether an image has been generated by artificial intelligence.
Google has launched its own marker, called SynthID, capable of identifying AI-generated content via metadata. Digimarc followed suit, creating a digital watermark for the same purposes. For its part, Microsoft has already integrated a watermark into content generated by Bing Image Creator. The problem is that with the proliferation of initiatives, it becomes difficult to navigate. The best would be if there was only one marker, used by everyone. Also, Adobe just revealed a “transparency icon” to mark all content produced by generative AI. Taking the form of a minimalist “cr” logo, for Content Credentialsit was developed by several member companies of the Coalition for Content Provenance and Authenticity (C2PA) group, including Adobe and Microsoft. “Given the sheer volume of content shared on social media platforms, Content Credentials provides consumers and media a reliable and transparent way to distinguish between what is actually owned or created by brands and what is duplicated, replicated or even falsified”, explains the company.
Content Credentials: a future universal marker?
Each time content is created or modified by generative AI, the “cr” symbol is placed on the image, video or PDF document. Simply click on it to display several pieces of information: its owner, its creator, its creation date, the AI used, etc. You can even access, through the Verify option, a detailed history of changes made over time. Of course, all of these elements are included in the metadata, so there is no point in erasing the photo icon to prevent access to the information.
Obviously, Content Credentials has been integrated into Adobe Firefly. But to be effective, the majority of companies using generative artificial intelligence would have to agree to adopt this logo. Microsoft has already agreed to abandon its own to put the little “cr” on images generated by Bing Image Creator and Microsoft Designer. The same is true for Publicis Groupe. As for Leica Camera and Nikon, they will affix this logo to photos taken by some of their devices. Adobe announced that in the coming months, other C2PA companies – which welcomes nearly 2,000 actors – should start using the logo. “We hope that the new icon will be so widely adopted that it will be universally anticipated and that it will one day become as ubiquitous and recognizable as the copyright symbol”, specifies Adobe. It remains to be seen whether Google will be up for it.