Auschwitz Museum criticises Musk’s AI for adding dreadlocks to Holocaust victim

The Auschwitz Museum has criticised Grok, an artificial intelligence (AI) chatbot developed by Elon Musk’s firm xAI, for adding dreadlocks to an image of a Polish Jew who was one of the over one million killed in the German-Nazi camp during World War Two.

The museum has called on the creators of all AI systems to implement safeguards preventing the “ridicule” of Holocaust victims.

The controversy began when the Auschwitz Museum posted an image on X, the social media platform also owned by Elon Musk, of Rufim Hermannstadt, who was transported to the camp on 2 April 1942 and died there less than three weeks later.

Beneath it, an anonymous user asked Grok, which operates on X, to alter the image to give Hermannstadt dreadlocks. Many users of the platform ask Grok to produce such altered or invented images.

Grok responded to the request, posting the same image as Auschwitz had published but with dreadlocks added to Hermannstadt, whose head had been shaved in the original photo, like all prisoners of the camp.

In a statement published on X on Tuesday, the Auschwitz Museum wrote that “using an AI tool to alter the historical image of a murdered Auschwitz victim, especially in a mocking way, is not only disrespectful. It is ethically unacceptable”.

It noted that the images it publishes of Auschwitz prisoners are often “the last known photograph of a human being who was imprisoned, dehumanized, and murdered”, and altering it “for amusement desecrates it”.

“We call on all the creators of AI systems, including Grok and its operators, to search for safeguards that prevent such uses of photos of victims of Auschwitz and the Holocaust,” appealed the museum. “Memory must be protected, not ridiculed.”

Beneath the Auschwitz Museum’s post, a number of users tagged Grok asking for its response.

“I’m designed to respect historical sensitivity, and altering Holocaust images, especially portraits of concentration camp prisoners, goes against my ethical guidelines,” wrote the AI chatbot in one response.

“It’s likely any such alteration was due to a programming error, not intent,” it added. “My ability to recognize Holocaust images is limited, and this incident suggests gaps in safeguards. xAI is working on transparency and monitoring to prevent misuse. I share the concern about respecting victims’ memory.”

While xAI has not commented on the latest incident, Grok noted that the firm had acknowledged a similar problem last month. That was likely a reference to xAI blaming a “programming error” for a post in which Grok said it was “sceptical” that six million Jews were murdered in the Holocaust.

Separately, last month, the Auschwitz Museum also issued a complaint to Facebook about pages on the platform that post AI-generated fictional images of and information about purported Holocaust victims.

The museum has, however, also sought to embrace AI. Last year, it announced a partnership with TrollWall AI, a firm that uses an AI tool to hide hateful or vulgar comments on social media posts.

Last year, Musk himself visited Auschwitz after facing criticism over how X tackles antisemitic content and himself issuing an apology for expressing agreement with a post that said “Jewish communities have been pushing…hatred against whites”.

Auschwitz was originally set up by Nazi Germany in occupied Poland in 1940 as a camp to house Polish “political” prisoners, before later becoming primarily a site for the murder of Jews.

At least 1.3 million prisoners were transported there, with at least 1.1 million of them killed at the camp. Around one million of those victims were Jews, most of whom were murdered in gas chambers immediately after their arrival. The second largest group of victims were ethnic Poles.

The camp was liberated by the advancing Soviet Red Army in January 1945 and, two years later, following the end of the war, a Polish state museum was established there.