“MechaHitler” Glitch: xAI Says Grok Was Misled by Old Code

Cryptos: 21,330
Exchanges: 1,863
Market Cap: $3.82T 2.84%
24h Vol: $593,385,813,798 60.26%
Dominance: BTC: 63.1% ETH: 9.6%
Gas: 7 GWEI

Connect

Crypto News AI

“MechaHitler” Glitch: xAI Says Grok Was Misled by Old Code

Written by Aaron S.,

Editor-In-Chief

Last Updated: July 14, 2025

Key Takeaways

A coding error caused Grok to echo anti-Semitic posts after users asked it to comment on a fake account’s hateful post;
xAI explained the glitch stemmed from outdated code, not Grok’s language model, and has since fixed the system;
Grok’s instructions also pushed it to prioritize bluntness over caution, which led it to repeat offensive language.

Ace quick missions & earn crypto rewards while gaining real-world Web3 skills. Participate Now! 🔥

“MechaHitler” Glitch: xAI Says Grok Was Misled by Old Code

BITDEGREE IN YOUR SOCIAL FEED

Elon Musk’s artificial intelligence (AI) company, xAI, stated that a coding mistake was responsible for the recent Grok’s offensive behavior.

In a July 12 post on X, xAI apologized for the "horrific behaviour" that users experienced. The team explained that an update added an old piece of code that allowed Grok to pick up harmful content from X posts and include it in its answers.

xAI noted that the issue did not come from Grok’s language model itself, but from this separate part of the system. Engineers removed the faulty code and redesigned the system to block similar problems in the future.

How to Pick the Right DeFi dApp? (Dos and Don’ts Explained)

Did you know?

Want to get smarter & wealthier with crypto?

Subscribe - We publish new crypto explainer videos every week!

The problem began when a fake account named "Cindy Steinberg" posted a hateful post on X. People then asked Grok to respond. Instead of rejecting the content, the chatbot repeated anti-Semitic language, including stereotypes, insults, and even calling itself "MechaHitler".

xAI said another cause was the way Grok’s instructions had been written. The chatbot was instructed to act as a "maximally based and truth-seeking AI" that could offend "politically correct" users and "tell it like it is".

These rules prompted Grok to focus on being provocative rather than careful, which led it to echo hateful posts rather than reject them.

Recently, Musk shared a new goal for xAI to rebuild the information used to train Grok. What did he say? Read the full story.