Meta’s AI Mislabeling: Real Photos Mistaken for Artificial Intelligence

MetaversePlanet June 24, 2024

0 1 minute read

Meta’s attempt to label AI-generated images on its platforms seems to be backfiring. While the company introduced the feature with good intentions in February, it’s facing criticism for mislabeling real photos and inconsistencies across platforms.

Many photographers and users are frustrated to find their genuine photos tagged as “Made with artificial intelligence” on Facebook, Instagram, and Threads. To add to the confusion, these labels appear inconsistently between the mobile app and web browser versions of the platforms.

Incorrect labels are a problem

Meta’s rollout of AI image labeling on Facebook, Instagram, and Threads has hit a snag. While the February launch aimed for transparency, photographers are up in arms about mislabeled real photos and inconsistencies across platforms.

The crux of the issue lies in Meta’s AI misinterpreting edits made with common software like Photoshop as AI manipulation. Photographers argue these edits shouldn’t trigger an “AI-made” label. Additionally, the inconsistency between mobile app and browser versions adds to the confusion.

Further clouding the issue, Meta remains silent on the specific implementation date of the labeling system and its inability to differentiate between basic edits and fully AI-generated content. To make matters worse, some clearly AI-generated images remain unlabeled on the platform.

Meta urgently needs to address these concerns. Refining their AI’s ability to distinguish edits from AI manipulation and ensuring consistent labeling across platforms are crucial steps. Transparency on the implementation timeline and a clear explanation of the labeling criteria would also be beneficial.