Meta has announced an update in its applications of the label that indicates that a content has been generated by artificial intelligence. The label, previously called “Made with AI”, will now be called “AI Information” (“AI Info”) and will appear on images supposedly generated or modified with AI.
This change comes after numerous complaints from many photographers, who claimed that the tag was being incorrectly applied to their photos. A notable example is that of former White House photographer Pete Souza, who saw that a photo of his from a basketball game taken 40 years ago had the tag, which could be due to the use of the Adobe Photoshop cropping tool.
“We are constantly improving our AI products and working closely with our industry partners on our AI labeling approach,” Meta spokesperson Kate McLaughlin told The Verge. According to the company, the new label aims to more accurately reflect that the content could be modified, rather than being completely generated by AI.
The problem lies in how platforms interpret the metadata applied by programs like Adobe Photoshop. After the expansion of Meta’s AI content labeling policies, many real images edited with Photoshop and published on Instagram, Facebook, and Threads started to be labeled as “Made with AI”.
Adobe, however, holds Meta responsible for how it presents that metadata. Andy Parsons, Senior Director of Adobe’s Content Authenticity Initiative (CAI), told The Verge that “when it comes to labeling AI, we believe platforms should only do so when an image is entirely AI-generated.”
The new labeling will be rolled out first in mobile apps and then in the web view, as explained by McLaughlin. When clicking on the new label, the same message as before will be displayed, with a detailed explanation of why it was applied and whether the image is completely generated by AI or edited with tools that include AI technology, such as Generative Fill.
Is it a success or a huge mistake? While the change can represent greater transparency by clarifying whether an image is completely generated by AI or simply “modified”, the new label could also “whitewash” the use of images that are indeed 100% generated by AI and give them a certain “apparent reality” for users, many of whom will only focus on the “Information about AI”.
