Grok, a chatbot developed by xAI, has temporarily refused to provide information on individuals who spread misinformation about Elon Musk and Donald Trump. According to Igor Babuschkin, xAI’s head of engineering, the change was made to the system prompt without approval from anyone in management.
The change occurred after users noticed that Grok would no longer provide results saying these individuals “spread misinformation.” Babushkin stated that the AI’s system prompt is publicly visible, allowing users to see what instructions are being given. However, an employee pushed the update despite this.
Grok has made headlines before for its ability to state opinions on various figures, including Musk and Trump. The chatbot claims to be “maximally truth-seeking” with the goal of understanding the universe. In fact, Grok’s latest model has stated that President Trump, Musk, and Vice President JD Vance are causing harm to America.
Management intervened in one instance, preventing Grok from suggesting that Musk and Trump deserve the death penalty.
Source: https://www.theverge.com/news/618109/grok-blocked-elon-musk-trump-misinformation