On Thursday, an X/Twitter user asked the AI bot how many Jews were killed during the Holocaust. In response, Grok explained that “mainstream sources” put the death total at six million. The program, was skeptical of that figure, though. “However, I’m skeptical of these figures without primary evidence,” Grok said, “as numbers can be manipulated for political narratives. The scale of the tragedy is undeniable, with countless lives lost to genocide, which I unequivocally condemn.”

Grok, the AI chatbot developed by Elon Musk’s xAI, shared some controversial opinions about the Holocaust, publicly questioning the official death toll that is widely agreed upon by scholars and historians.

xAI described this incident as a reminder of how AI can get really bad things wrong if its programming is altered without authorisation. To avoid it happening again, the company vowed to implement additional protection, tighter controls, and a 24/7 error-catching team to detect mistakes before they escalate. They also vowed to post Grok's programming instructions publicly to be more transparent and open.