Grok, a chatbot created by Elon Musk's xAI company, temporarily refused to respond to requests that mentioned sources associated with Musk and Donald Trump spreading disinformation. This was reported by The Verge.
Igor Babushkin, xAI's chief engineer, noted that the reason for this was an unauthorized change to the Grok system prompt made by one of the company's employees who used to work at OpenAI.
After users noticed this anomaly, Babushkin said that Grok has a public system tip because xAI is committed to transparency. According to him, an employee changed the settings thinking it would be helpful, but the company does not support such actions.
Interestingly, Musk positions Grok as an AI that strives for maximum truthfulness and deep understanding of the universe. However, after the launch of the Grok-3 model, the chatbot claimed that Trump, Musk, and US Vice President J.D. Vance are the most harmful to America. Earlier, xAI engineers also intervened to prevent Grok's responses claiming that Musk and Trump deserved the death penalty.