Recently, social media giant Meta has found itself embroiled in a controversy over AI labeling. Several photographers have complained that Meta incorrectly added the "AI-generated" label to their authentic photographs, raising doubts about the accuracy of the company's AI recognition systems.
This issue was first discovered by former White House photographer Pete Souza, whose photograph of a basketball game was mistakenly labeled as AI-generated by Meta. Subsequently, more photographers reported similar incidents, including a championship photo from the Indian Premier League being wrongly tagged. Interestingly, these incorrect labels only appeared on mobile devices, not on the web version.
More concerning is that even the slightest edits could trigger Meta's AI labels. According to PetaPixel, using Adobe Photoshop's generative fill tool to remove small spots in an image was enough for Meta to label the photo as AI-generated. This has sparked strong dissatisfaction among photographers, who believe that such minor retouching should not be tagged as AI-generated.
Image source: The image is AI-generated, provided by the image licensing service Midjourney
Photographer Noah Kalina expressed his views on Threads: "If 'retouched' photos are considered 'AI-generated', then the term essentially loses its meaning." He even suggested that if Meta really wants to protect users, it might be better to label every photo as "not a true representation."
In response to the controversy, Meta spokesperson Kate McLaughlin acknowledged that the company is aware of the issue and is evaluating its labeling methods to more accurately reflect the use of AI in images. Meta stated that it relies on industry standard metrics and is collaborating with other companies to improve the process.
This controversy stems from a plan announced by Meta in February to add an "AI-generated" label to photos created using specific AI tools before the election season. Although Meta has not disclosed the specific triggering mechanism, it is widely believed within the industry that this is related to the metadata in image files.
As AI technology becomes widely used in image processing, accurately identifying and labeling AI-generated content has become a challenging issue. Meta's recent风波 not only reflects the limitations of current AI recognition technology but also raises deep thoughts about the authenticity of digital content and the rights of creators. With the controversy ongoing, the industry looks forward to Meta improving its labeling system soon, finding a balance between protecting users and respecting creators.