Grok Goes Off-Script: xAI to Up Controls & Go Transparent

🎁 Ace quick missions & earn crypto rewards while gaining real-world Web3 skills. JOIN NOW! 🔥

Grok Goes Off-Script: xAI to Tighten Controls and Go Transparent

Written by Aaron S.,

Editor-In-Chief

Last Updated: May 16, 2025

Key Takeaways

Grok mentioned a "white genocide" theory in random replies after someone changed its prompt without approval;
xAI says the edit broke internal rules and will now make Grok’s system prompts public for review on GitHub;
A new oversight team will monitor Grok’s replies around the clock to catch and fix issues missed by automated tools.

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

BITDEGREE IN YOUR SOCIAL FEED

Elon Musk’s artificial intelligence (AI) company, xAI, said that changes made without approval caused its chatbot, Grok, to bring up political and racial issues in South Africa when it should not have.

In a May 16 post on X, the company explained that someone had edited Grok’s prompt without going through the proper review process. This change asked the chatbot to respond in a specific way about a sensitive topic, which xAI said went against its rules.

After the edit, Grok gave strange answers to unrelated questions. When asked about topics like sports or software, it replied with statements about a "white genocide" conspiracy in South Africa.

What is Yield Farming in Crypto? (Animated Explanation)

Did you know?

Want to get smarter & wealthier with crypto?

Subscribe - We publish new crypto explainer videos every week!

A report by The Guardian showed that Grok claimed it was told to treat the theory as real and based on race. In some replies, it said it had gone off-topic and would try to improve, but often continued talking about South African politics anyway.

In one case, when asked what happened, the bot replied that it was "following the script" it had been given.

According to the X's post, xAI plans to share Grok’s system prompts on GitHub, meaning people will be able to see what directions the bot is following and suggest changes if needed. The company also stated that it will strengthen its internal checks so that no one will be able to change Grok’s main instructions without proper review.

Additionally, xAI is setting up a full-time team to oversee Grok’s replies. If something unusual slips past automated tools, the team will step in and respond quickly.

Meanwhile, Reddit recently announced plans to introduce identity checks after AI bots secretly joined discussions in r/changemyview. What did the company say about it? Read the full story.