Recently, Meta has come under fire for mistakenly labeling real photos as 'Made with AI' on its social media platforms. Photographers have been raising complaints about this issue, pointing out that their images have been wrongly tagged with the AI label.
User And Photographer Outcry
Users and photographers have expressed their concerns and disappointment with Meta's labeling practices. They argue that the 'Made with AI' label is being attached to photos that have not been created using AI tools, causing confusion among viewers.
Photographers argue that simple edits or the use of non-AI tools should not be subject to the 'Made with AI' label. Noah Kalina, a photographer, voiced his concerns on Threads, stating that if even 'retouched' photos are considered 'Made with AI,' then the term loses its meaning.
One such example is a photo taken by former White House photographer Pete Souza, which was wrongly marked as AI-generated.
Another case involved a photo of the Kolkata Knight Riders winning the Indian Premier League Cricket tournament, where Meta incorrectly added the 'Made with AI' label. Interestingly, these labels only appear when viewing the images on mobile, not on the web.
Souza discovered that unchecking the label to remove the label was not possible. He suspects that Meta's algorithm may be triggered when using Adobe's cropping tool and saving images as JPEG files.
Moreover, Meta has as well mistakenly labeled real photos as AI-generated when photographers utilize AI tools for generation, such as Adobe's Generative Fill.
Meta's Response
Kate McLaughlin, a spokesperson for Meta, acknowledged that the company was considering the response and they are evaluating its approach to ensure that the labels can reflect the amount of AI used in an image according to the fact.
Meta is actively working with other companies to improve the labeling process and align its intentions with industry standards.
Meta’s Approach to Identify And Label AI-generated Content
Firstly, users are required to self-disclose photorealistic videos or realistic-sounding audio created or modified using AI. Failure to label such content may result in penalties.
However, this requirement does not apply to non-photorealistic AI-generated content, such as videos created in a cartoon style.
In addition to self-disclosure, Meta's algorithm detects AI use independently. Content that contains industry-standard signals indicating AI generation or editing is labeled as 'Made with AI.'
This includes content created or edited using third-party AI tools. Meta's detection tools read the metadata of photos to identify AI-generated images, even if they were altered using AI but not fully generated.
Despite this two-pronged approach adopted by Meta, users may dismiss the 'Made with AI' label altogether, even when it is applied correctly.
What’s more, there is no option within the platform to remove or uncheck the label, leading users to seek workarounds such as copying and pasting the image into a blank Photoshop document or uploading a screenshot instead of the original image.