A Facebook study is reportedly being flooded with bizarre AI-generated images of Jesus, flight attendants, muddy sharks, and more without any AI watermarks, despite the company's requirement that users label AI-generated images.
The viral AI images have raised concerns about the application's AI safeguards against scammers and spammers behind the images.
Most of the pictures are from dozens of Facebook groups that update almost every hour of the day; strangely, many of these pages are themed around Jesus and flight attendants.
Some of those pages have amassed substantial followings in recent months, and the posts frequently elicit comments that range from flattering compliments akin to those left by bots to sarcastic remarks directed at people who seem to think the photos are real.
(Photo: KIRILL KUDRYAVTSEV/AFP via Getty Images) Meta is removing the news tab feature on its Facebook social media platform in the US and Australia
The study claims that spammers and scammers, who appear to be driven more by money or power than by ideology, are responsible for these AI-generated graphics. According to reports, their pages are leveraging AI-generated photos to get a lot of traction on Facebook.
It is more troubling that the Facebook Feed occasionally suggests unlabeled AI-generated images to users who do not follow the Pages that are posting them or are unaware that the images are AI-generated. This emphasizes the need for increased transparency and provenance standards as AI models become more common.
Viral AI Images
120 Facebook Pages with at least 50 AI-generated photos each were examined in the study, and the images were categorized as Pages classified as "other creator," "spam," and "scam." A few were synchronized groups of Pages that identical administrators operated. The Pages had an average of 128,877 followers as of March 5, 2024, and a median of 71,000 followers.
These AI photos generated a total of hundreds of millions of interactions and exposures. One of the top 20 most-viewed posts on Facebook in Q3 2023 featured an AI-generated image, garnering 40 million views and over 1.9 million interactions.
The pictures range from detailed artistic representations of Jesus made from sand or plastic bottles to Jesus recently in the form of sea animals, primarily shrimp, crabs, and seahorses. Some of the posts even assert that the images were created with their "own hands."
Meta's AI Labelling Efforts
While Meta requires users to mark AI-produced content on its platforms while the tech giant works toward methods to automatically recognize such content, the great majority of these synthetic photos do not make it clear in the post or on the page that they are AI-created content.
The method the company uses to distinguish AI-generated artwork combines visible indicators with watermarks included in image files but invisible to the naked eye.
Meta, in the past, said it wanted to improve the accuracy of content tagging by enhancing its AI systems' detection capabilities through industry partnerships and shared technological standards, but the tech giant has yet to comment on the viral AI images.
Related Article: Study Finds Top AI Image Generators are Still Producing Inaccurate Election Images Despite Pledges
