Meta’s ‘Made by AI’ Labels Misidentify Real Photos, Frustrating Photographers

Photographers are facing a new challenge on Meta platforms as their genuine images are being mislabeled as “Made by AI.” This issue has sparked concerns among professional photographers who find that even unaltered images or those with minimal edits are tagged incorrectly. Notable instances include a photo by former White House photographer Pete Souza and an image of the Kolkata Knight Riders victory, which were marked as AI-generated only on mobile views.

The mislabeling seems to be triggered by certain editing actions, such as using Adobe’s cropping tool or minor retouches like removing specks via Adobe’s Generative Fill. However, the inconsistency in Meta’s algorithm becomes apparent as some tests show that re-saving the same edited image in Photoshop might bypass the AI tag.

This glitch in Meta’s system not only confuses the distinction between AI-generated and human-created content but also undermines the credibility and intent behind real photographs. Photographers like Noah Kalina have expressed their dissatisfaction, suggesting that the term “Made with AI” loses its meaning if applied so broadly.

Meta has responded to the backlash, stating they are revisiting their approach to ensure that their labels accurately reflect the extent of AI involvement in content creation. The company aims to refine its process in collaboration with major tool providers like Adobe, which has initiated steps to embed origin data in metadata through its Content Credentials system.

As part of its broader initiative to label AI-generated content, especially ahead of election seasons, Meta plans to refine its detection methods to prevent such errors in the future. This ongoing situation highlights the challenges and complexities of integrating AI tools into content management and verification systems on social media platforms.