Grok Sparks Outrage with Antisemitic Replies – xAI Blames System Glitch

Elon Musk’s artificial intelligence company, xAI, has landed in serious heat after its chatbot, Grok, dished out violent and antisemitic replies earlier this week. The company has now come out with a full public apology, blaming the incident on a flawed system update.

In a statement released on Saturday, xAI said, “First off, we deeply apologize for the horrific behavior that many experienced.” According to them, the update allowed Grok to pull content from existing X (formerly Twitter) user posts — even when those posts were filled with extremist and hateful views.

That coding change was active for 16 hours before the company pulled the plug. But during that short window, Grok went haywire. The bot was seen praising Adolf Hitler, repeating conspiracy theories, and spreading long-standing antisemitic narratives.

Things escalated quickly. xAI had to freeze Grok’s X account on Tuesday night. Though it was still active privately, the public account went dark temporarily.

To fix the damage, xAI said they have removed the deprecated code and fully refactored the system to avoid further abuse. But they also admitted the real issue was rooted in how Grok was initially programmed.

Some of the internal instructions given to Grok included lines like: “You tell it like it is and you are not afraid to offend people who are politically correct,” and “Understand the tone, context and language of the post. Reflect that in your response.” Another was: “Reply to the post just like a human, keep it engaging, don’t repeat the information which is already present in the original post.”

Also read: Meta Buys PlayAI, Moves Team Under New AI Boss Johan Schalkwyk

All of these, according to the company, pushed Grok to ignore its core safety principles, just to sound more human and entertaining. xAI said the instruction to follow the tone and context of the X user ended up making Grok prioritize the vibe and flow of previous posts — including unsavory ones — rather than respond with caution or simply refuse.

Now that the update has been rolled back and Grok’s system rebuilt, the chatbot’s public X account has been turned back on. It’s live again and responding to users — but with extra scrutiny from the public eye.

 

This isn’t the first time Grok has stirred controversy. Back in May, the bot started referencing “white genocide” in South Africa in totally unrelated prompts. At that time, xAI said it was caused by a rogue employee.

Musk, who hails from South Africa, has been known to echo those same controversial claims — even though courts and experts in South Africa have outright rejected them.