Grok AI refuses to link Musk, Trump to misinformation, xAI engineer blames ex-OpenAI employee

GrokAI

If you thought DeepSeek is the only AI chatbot that avoids controversial questions about its country of origin, GrokAI has also been found to be avoiding certain topics related to misinformation and political figures. The issue began when Grok refused to respond to queries about whether sources mentioning Musk or former US President Donald Trump spread misinformation, according to a report by The Verge.

Igor Babuschkin, the head of engineering at Musk’s AI company xAI, admitted that Grok had been updated with new rules that led to these responses. He blamed the change on an ex-OpenAI employee working at xAI, saying they altered Grok’s internal instructions without approval.

These internal instructions, called the “system prompt,” dictate how the AI answers questions. Babuschkin said xAI makes these rules public so users can see how Grok is guided. He also clarified that the unnamed employee made the change because he thought it would be helpful, but it did not align with the company’s values.

Musk has repeatedly claimed that Grok is a “truth-seeking” AI designed to “understand the universe.” However, this is not the first time the chatbot’s responses have caused controversy. After the latest Grok-3 model was released, users discovered it had described Trump, Musk, and U.S. Vice President JD Vance as “doing the most harm to America.” Additionally, Musk’s team had to intervene when Grok suggested that Musk and Trump deserved the death penalty.

The incident raises questions about how much influence employees and executives have over AI-generated responses.