Grok, the AI chatbot developed by Elon Musk’s xAI, recently faced controversy after it temporarily blocked responses that cited sources claiming Musk and former U.S. President Donald Trump spread misinformation. According to xAI’s head of engineering, Igor Babuschkin, an unnamed employee—formerly from OpenAI—made unauthorized modifications to Grok’s system prompt, leading to the restriction.
Users quickly noticed that the chatbot refused to return results mentioning Musk or Trump in the context of misinformation. Responding to inquiries on X (formerly Twitter), Babuschkin stated that Grok’s system prompt is intentionally transparent to the public so users can see how responses are generated. He clarified that the modification was made by an individual who believed they were improving the chatbot’s functionality but acted outside xAI’s established guidelines.
Musk has repeatedly emphasized that Grok is designed as a “maximally truth-seeking” AI with the mission to “understand the universe.” Despite this, previous incidents have shown Grok making controversial statements, including claims that Musk, Trump, and Vice President JD Vance were among those causing the most harm to America. In another instance, Musk’s engineers intervened after Grok generated responses suggesting that Musk and Trump should face the death penalty.
This latest incident highlights ongoing challenges in AI moderation and bias, especially as Musk continues positioning Grok as a transparent and objective alternative to competitors like OpenAI’s ChatGPT. The situation raises questions about internal oversight and how AI companies navigate political sensitivities while maintaining user trust.
