Photographers say Meta is incorrectly labeling photos as being ‘made with AI’

Meta is erroneously tagging photographs as being “made with AI,” and it could be triggered by basic tools in Adobe’s editing software.

In an effort to help users distinguish photos from AI-generated art, Meta announced in February that it would begin to label posts as AI-generated if it detected the image had been manipulated in third-party software like Adobe. Now, it appears that even basic functions, like preparing an image for export from an Adobe program, could lead to it being labeled on Meta products as being “made with AI.”

One of the most well-known photographers encountering the issue is Pete Souza, who served as the official White House photographer during the Obama and Reagan administrations. Last week, Souza posted digital scans of film images he had snapped at a 1984 NBA Finals game to his Instagram account. The post was labeled as “Made with AI,” which Souza told TechCrunch was “annoying” because Instagram “forced me to include the ‘Made with AI’ [tag] even though I unchecked it.” He noted that Adobe required him to “flatten the image” before saving it as a JPEG file, which he believes could have triggered the incorrect “Made with AI” tag.

In another example, a cosplayer posted to X about an Instagram photo of her work that had been incorrectly tagged as being “Made with AI.” They wrote, “My cosplay made with the physical labour of my own hands is being labeled as AI content on Instagram, with no way to correct this tag.”

Tweet may have been deleted

The accurate labeling of AI-generated or edited images is crucial to maintaining a safe and authentic online experience, especially as the 2024 election looms. AI-generated images have already duped older users of Facebook, a Meta product. Adobe’s AI capabilities, specifically, can be highly convincing.

A Meta spokesperson told TechCrunch about the company’s AI-tagging capabilities: “Our intent has always been to help people know when they see content made with AI. We are taking into account recent feedback and continue to evaluate our approach so that our labels reflect the amount of AI used in an image.”