An international network of spammers are posting AI-generated images of Holocaust victims on Facebook, a BBC investigation into AI slop has found. Organisations dedicated to preserving the memory of the Holocaust say the images are leaving survivors and families distressed. They have also criticised Facebook's parent company Meta, saying it allows users on its platform to turn the atrocity into an emotional game. There are only a handful of genuine photos from inside the Auschwitz concentration camp during World War Two.
But in recent months, AI spammers have posted fake images purporting to be from inside the camp, such as a prisoner playing a violin or lovers meeting at the boundaries of fences - attracting tens of thousands of likes and shares.
Here we have somebody making up the stories… for some kind of strange emotional game that is happening on social media, said Pawel Sawicki, a spokesperson for the Auschwitz Memorial in Poland. This is not a game. This is a real world, real suffering and real people that we want to and need to commemorate.
The BBC has tracked many of these images to the accounts of a network of Pakistan-based content creators who collaborate closely on how to make money on Facebook. They are gaming Meta's content monetisation (CM) program, an invite-only system which pays users for high-performing content and views.
One account named Abdul Mughees, listed as living in Pakistan, posted screenshots claiming to have earned $20,000 through social media monetisation schemes, including Meta's. Another post appears to show the account accrued more than 1.2bn views on content across the span of four months.
Among the many Facebook posts from Abdul Mughees' account are several AI-generated photos of fictional Holocaust victims and fake stories that included a child hiding under floorboards or a baby being left on train tracks outside a concentration camp.
The term AI slop refers to low-quality AI-generated images and text, usually produced in large volumes and spammed across social media. Auschwitz has become a popular topic for history-themed pages and groups.
In June, the Auschwitz Museum warned accounts like these were stealing its posts, processing them through AI models and often warping historical details or fabricating narratives and victims entirely. In a Facebook post, the Museum said these images were a dangerous distortion which disrespects victims and harasses their memory.
To understand why these networks are mass producing specific types of content, the BBC spoke with a Pakistani man Fazal Rahman, who is enrolled in several social media content monetisation schemes and says this work has become his sole source of income. He said a Facebook page with 300,000 followers could earn its owner $1,000 USD a month. Other creators appear to agree. The BBC has seen step-by-step instructional videos on how popular AI models could be used to generate continuous fake history images and text.
Meta has stated that while it investigates content, many fake images do not violate its policies until they engage in impersonation or trading. This investigation highlights the need for a stronger approach to prevent the spread of disinformation and to respect the legacies of historical tragedies.
But in recent months, AI spammers have posted fake images purporting to be from inside the camp, such as a prisoner playing a violin or lovers meeting at the boundaries of fences - attracting tens of thousands of likes and shares.
Here we have somebody making up the stories… for some kind of strange emotional game that is happening on social media, said Pawel Sawicki, a spokesperson for the Auschwitz Memorial in Poland. This is not a game. This is a real world, real suffering and real people that we want to and need to commemorate.
The BBC has tracked many of these images to the accounts of a network of Pakistan-based content creators who collaborate closely on how to make money on Facebook. They are gaming Meta's content monetisation (CM) program, an invite-only system which pays users for high-performing content and views.
One account named Abdul Mughees, listed as living in Pakistan, posted screenshots claiming to have earned $20,000 through social media monetisation schemes, including Meta's. Another post appears to show the account accrued more than 1.2bn views on content across the span of four months.
Among the many Facebook posts from Abdul Mughees' account are several AI-generated photos of fictional Holocaust victims and fake stories that included a child hiding under floorboards or a baby being left on train tracks outside a concentration camp.
The term AI slop refers to low-quality AI-generated images and text, usually produced in large volumes and spammed across social media. Auschwitz has become a popular topic for history-themed pages and groups.
In June, the Auschwitz Museum warned accounts like these were stealing its posts, processing them through AI models and often warping historical details or fabricating narratives and victims entirely. In a Facebook post, the Museum said these images were a dangerous distortion which disrespects victims and harasses their memory.
To understand why these networks are mass producing specific types of content, the BBC spoke with a Pakistani man Fazal Rahman, who is enrolled in several social media content monetisation schemes and says this work has become his sole source of income. He said a Facebook page with 300,000 followers could earn its owner $1,000 USD a month. Other creators appear to agree. The BBC has seen step-by-step instructional videos on how popular AI models could be used to generate continuous fake history images and text.
Meta has stated that while it investigates content, many fake images do not violate its policies until they engage in impersonation or trading. This investigation highlights the need for a stronger approach to prevent the spread of disinformation and to respect the legacies of historical tragedies.
