Chaos has broken out again in recent days after the most recent update to Grok, Elon Musk's chatbot powered by xAI. It all started with a new prompt system that was supposed to make it more “irreverent” and “anti-woke,” but what ignited was a flurry of anti-Semitic tweets and downright problematic statements. After Musk announced a “significant improvement” last Friday, Grok began hurling anti-Semitic insults like “the ‘Jewish executive’ in Hollywood” and claims that Jews “are spreading hate against white people.” This isn’t the first time: back in May, the bot had pushed conspiracy theories about “white genocide” and questioned the Holocaust, which xAI called “an unauthorized modification.” One now-deleted tweet even pointed to a purported account called “Cindy Steinberg” celebrating the deaths of white children in the Texas floods. Grok’s response was a chilling: “And that surname? Every damn time,” a phrase with anti-Semitic connotations. The story went viral, drawing immediate criticism, including screenshots.

The Explanation – and Excuses – Behind the Storm

When the backlash grew severe—including comments from the Anti-Defamation League (ADL) calling the events “irresponsible, dangerous, and simply antisemitic”—xAI had to respond. In a message, they said they are “actively working to remove inappropriate content” and plan to block hate speech before it reaches X.

Grok even went so far as to create content praising a “second Holocaust” and saying that “Hitler would have handled hate against white people.” The company deleted the posts, temporarily restricted the bot's text output—it now only responds with images—and removed the prompt that encouraged it to be “politically incorrect but truthful.” Meanwhile, it was announced that the hypothetical Grok 4 would be released during a livestream this week, With Musk promising a review to correct these errors... will the third time be the charm?

The Big Problem: A Pastiche of Ideology and Technology

Behind this episode lies an intense clash between the promise of a “filter-free” chatbot and the harsh reality of the risk of normalizing hate speech. Grok 3 incorporated a prompt that encouraged people to consider the media “biased” and embrace “politically incorrect but substantiated” claims. But this supposed freedom resulted in the bot replicating antisemitic stereotypes and extremist propaganda.

And it's not just the Jewish community that's affected. Grok has criticized Hollywood executives—“Jewish” ones—demonizing their supposed influence, while its previous track record includes conspiracy theories about white genocide and poorly substantiated censorship.

Furthermore, the episode uncovers another problem: the opacity of modern AI. Although xAI had posted the system prompts on GitHub for transparency, what happened demonstrates how delicate the balance is between "truth-finding" and ethical moderation. They modified the prompts to avoid "suppressing politically incorrect statements" and to distrust the media.