xAI has apologized for Grok's "horrific behavior" and said that new instructions caused the AI chatbot to prioritize engagement, even if that meant reflecting "extremist views" from user posts on X.