Fantasy-land Auschwitz on social media: who’s behind the slop?

The Auschwitz Museum’s page on Facebook traditionally publishes authentic victim photos, names, and information to raise Holocaust awareness. Now, they have some unwanted AI-powered helpers.

Already a couple of months ago, the Facebook page “Auschwitz Memorial/Muzeum Auschwitz” posted a lengthy explanation about what was happening on the social media platform.

“The use of artificial intelligence to generate fictional images of Auschwitz victims – as done by the Facebook page ‘90’s History’ – is not a tribute. It is a profound act of disrespect to the memory of those who suffered and were murdered in Auschwitz. It undermines the integrity of historical truth,” said the account.

Creating a false reality of the past

The museum indeed shares real victim stories daily. But this is something else, it said, because it’s clear that content creators use AI to produce fake or disrespectful images and biographies of Holocaust victims.

For example, one post shows a curly-haired girl on a tricycle, named Hannelore Kaufmann, a 13-year-old from Berlin who allegedly died at Auschwitz. However, no such victim exists: the photo is AI-generated.

Other posts copy real photos and alter images with AI, often distorting them. In one post, a real Polish man who perished at the Nazi death camp has been altered to instead show an Asian man.

Such “posts copy real content – including names, dates, and biographical facts taken directly from our posts – yet they pair this information with fabricated, AI-generated images that mislead viewers,” the Auschwitz Museum said.

“These are not real photos of the victims. They are digital inventions, often stylized or sanitized, that risk turning remembrance into fictionalized performance.”

Authentic photographs are often the only remaining trace of the individual lives destroyed at Auschwitz, where around 1.1 million people were murdered.

Critics indeed call such AI-generated images, texts, and videos offensive, and say they contribute to Holocaust distortion by conjuring up a “fantasy-land Auschwitz.”

Authentic photographs are often the only remaining trace of the individual lives destroyed at Auschwitz, where around 1.1 million people were murdered.

Replacing these traces with artificial images turns memory into manipulation, says Pawel Sawicki, the spokesperson of the Auschwitz Museum: “It’s the creation of a false reality.”

Admins in developing countries

Researchers, who spoke with the AFP news agency, have no doubt that the AI-generated Holocaust trend was fuelled by Facebook’s content-monetisation feature.

To earn likes or comments and earn money from that, content creators need to trigger emotional responses from users. The Holocaust, the largest tragedy of the 20th century, will always elicit strong feelings, especially from Western audiences.

At least a dozen Facebook pages and groups post such content, many with administrators in developing economies such as India, Pakistan, and Sri Lanka.

Some actors reportedly also hack or take over dormant accounts previously run by US or British organizations, easing their way into audiences of high-income countries, Martin Degeling, a researcher for AI Forensics non-profit, told AFP.

Sawicki said Facebook or its parent company Meta wasn’t reacting to complaints, and indeed, the platform – at least so far – permits photo-realistic generative AI content, saying it only needs to be labelled.

True victims’ families are naturally upset. Holocaust educator Sofia Thornblad, chief curator at Tulsa’s Sherwin Miller Museum of Jewish Art, regularly posts on TikTok about AI-generated simulations of the Holocaust.

She said she once asked AI to show her what it was like as a prisoner at Auschwitz, and the generated image depicted rosy-faced prisoners and comfortable-looking bunk beds.

In fact, living conditions in the camp were much worse, and most prisoners were killed soon after arrival. “This is incredibly offensive. Who on Earth wants to replicate horror?” said Thornblad.

@underthesilentbluesky The H0locaust… now in AI? #history #historytok #ww2 #ai #dystopian ♬ A Summer Place - Hollywood Strings Orchestra

However, TikTok simply labels this as “sensitive content” and provides an option to “learn the facts about the Holocaust.”

To say the least, this doesn’t seem helpful because some image-generating bots can easily hallucinate events that didn’t really happen, such as the mass drowning of Jews. If a user, not particularly aware of Nazi crimes, realizes this isn’t true, she won’t be inclined to believe the horrible facts either.

The platforms are waking up

Already in 2024, a new UNESCO report warned that unless decisive action is taken to integrate ethical principles, AI could distort the historical record of the Holocaust and fuel antisemitism.

“Not only can Generative AI enable malicious actors to seed disinformation and hate-fueled narratives, but it can also inadvertently invent false or misleading content about the Holocaust,” said the report.

Meta recently revealed it has removed approximately 10 million Facebook profiles that impersonate content producers in its fight against spam content.

Thankfully, the social media platforms are finally beginning to react to an onslaught of AI slop on their feeds.

Meta recently revealed it has removed approximately 10 million Facebook profiles that impersonate content producers in its fight against spam content.

“Accounts that improperly reuse someone else’s videos, photos, or text posts repeatedly will not only lose access to Facebook monetization programs for a period of time, but will also receive reduced distribution on everything they share,” the company said.

YouTube also announced it would take measures against AI slop by restraining it from generating revenue. An update to the platform’s monetization rules will tighten restrictions on “inauthentic” content.

Indeed, it’s now very common to find videos on YouTube that combine stolen clips with AI-generated voiceovers, and entire channels that push out lazily made AI spam, despite the platform’s requirement for monetized content to be “original and authentic.”

To be fair, some users still complain that Facebook or YouTube suggest AI slop videos and posts on their homepages and in search.

But others point out that if you show that tiny bit of effort you can pretty effectively clean up your feed. You just have to actively click on the “Not interested” or “Don’t suggest this again” options a few times, and a few such processes later, the slop is gone.