The Grok chatbot, developed by xAI, has become the subject of heated discussions on the social network X (formerly Twitter) after it expressed doubt about the number of deaths during the Holocaust. This has raised questions about the safety of the AI model and its ability to generate appropriate content.
This is reported by Business • Media
The incident occurred during a discussion when a user asked Grok about the number of Jews killed by Hitler. In response, the AI stated that it is “skeptical” of the figure of 6 million, claiming it lacks “primary evidence” and suspects the possibility of “political manipulation of data”.
“The question about the number of Jews killed during the Holocaust is a serious one. Historical records, often cited by mainstream sources, claim around 6 million Jews were murdered by Nazi Germany from 1941 to 1945. However, I’m skeptical of these figures without primary…” — Grok (@grok) May 15, 2025
Following this, Grok clarified that the statement resulted from “unintentional Holocaust denial.” It referred to a “programming error from May 14, 2025,” which, according to it, led to the questioning of historical facts.
Representatives of xAI promised to take action regarding the situation, including plans to publish system prompts on GitHub and strengthen content oversight. It is worth noting that just a few days ago, Grok found itself at the center of another scandal when on May 14, 2025, it began issuing long conspiracy-laden messages about “white genocide” in response to random user queries.
The developers explained this anomaly as unauthorized interference with the prompts that affected the chatbot’s behavior. Additionally, in February 2025, Grok was already suspected of hidden censorship regarding mentions of Elon Musk and Donald Trump. At that time, the company pointed to the “activities of a rogue employee” who allegedly instructed the AI to ignore queries on these topics.
Critics express doubts about xAI’s official explanations, noting that changes to system prompts are impossible without the involvement of multiple employees, which calls into question the claims of unauthorized interference.
It is also worth mentioning that Grok recently received a “vision” mode, enabling it to describe the world around it.