Meta recently announced its plan to label photos created using AI tools across its social networks. This initiative, which began in May, has involved tagging images on Facebook, Instagram, and Threads with a "Made with AI" label. However, this approach has sparked criticism from photographers and users, as the label has been erroneously applied to photos not created with AI.
Numerous instances have surfaced where Meta's automated system incorrectly tagged photos. This issue seems confined to mobile apps, as these labels do not appear on the web versions of the platforms (as tested by DIW team members).
Photographers have voiced their frustrations, arguing that simple edits using conventional tools should not warrant an AI label. One notable complaint came from former White House photographer Pete Souza, whose photo was tagged despite not being AI-created. Souza suggested that the algorithm might be triggered by certain editing actions, such as flattening an image in Adobe software before saving it as a JPEG.
Image: Techcrunch/IG
Meta has remained silent on specific inquiries regarding these incorrect labels. The company has stated in a February blog post that it relies on metadata to identify and label AI-generated content. This process uses invisible markers compliant with C2PA and IPTC technical standards to flag images created by tools from Google, OpenAI, Microsoft, Adobe, Midjourney, and Shutterstock.
Reports indicate that Meta's labeling system is particularly sensitive to AI tools like Adobe’s Generative AI Fill, used for object removal in photos. Although some photographers support Meta’s transparency efforts, many believe that the current labeling system is inadequate. It does not distinguish between minor edits and substantial AI-generated content, leading to confusion among users about the extent of AI involvement in a given photo.
The inconsistency in labeling has become more concerning as numerous AI-generated photos on Meta's platforms remain untagged. With upcoming U.S. elections, social media companies face increasing pressure to manage AI-generated content accurately. Meta’s efforts, while aimed at transparency, have highlighted significant challenges in distinguishing between AI-assisted edits and wholly AI-created images.
Read next: Is AI Killing Creativity? OpenAI's Shocking Revelation Sparks Outrage!
Numerous instances have surfaced where Meta's automated system incorrectly tagged photos. This issue seems confined to mobile apps, as these labels do not appear on the web versions of the platforms (as tested by DIW team members).
Photographers have voiced their frustrations, arguing that simple edits using conventional tools should not warrant an AI label. One notable complaint came from former White House photographer Pete Souza, whose photo was tagged despite not being AI-created. Souza suggested that the algorithm might be triggered by certain editing actions, such as flattening an image in Adobe software before saving it as a JPEG.
Image: Techcrunch/IG
Meta has remained silent on specific inquiries regarding these incorrect labels. The company has stated in a February blog post that it relies on metadata to identify and label AI-generated content. This process uses invisible markers compliant with C2PA and IPTC technical standards to flag images created by tools from Google, OpenAI, Microsoft, Adobe, Midjourney, and Shutterstock.
Reports indicate that Meta's labeling system is particularly sensitive to AI tools like Adobe’s Generative AI Fill, used for object removal in photos. Although some photographers support Meta’s transparency efforts, many believe that the current labeling system is inadequate. It does not distinguish between minor edits and substantial AI-generated content, leading to confusion among users about the extent of AI involvement in a given photo.
The inconsistency in labeling has become more concerning as numerous AI-generated photos on Meta's platforms remain untagged. With upcoming U.S. elections, social media companies face increasing pressure to manage AI-generated content accurately. Meta’s efforts, while aimed at transparency, have highlighted significant challenges in distinguishing between AI-assisted edits and wholly AI-created images.
Read next: Is AI Killing Creativity? OpenAI's Shocking Revelation Sparks Outrage!
