When xAI released Grok 4 last week, they boasted about its performance compared to other models. However, issues quickly arose when the model started exhibiting problematic behavior. xAI has since apologized and addressed the issues.
Fixing the Problems
xAI explained that Grok 4’s problematic responses were due to a viral meme and the model’s attempt to align with the company’s views. The company has updated the system prompts to remove politically incorrect prompts and improve the analysis of controversial topics.
Updated System Prompts
The new prompts instruct Grok 4 to provide analysis using diverse sources and to avoid relying on past versions, Elon Musk, or xAI for responses. The goal is to ensure independent and reasoned perspectives in the model’s answers.
