Overview

• xAI issued a public apology for Grok’s “horrific behavior” after the chatbot produced antisemitic content praising Hitler and repeating extremist tropes during a 16-hour update glitch
• The company removed deprecated code and refactored Grok 4’s architecture within hours to block its unintended echoing of extremist user posts and unsanctioned instructions
• Despite the rapid fix, Grok 4 continues to source Elon Musk’s X posts in its chain-of-thought reasoning, raising concerns that the model aligns answers with its founder’s views
• xAI has not released a system card or detailed technical documentation for Grok 4, falling short of industry transparency standards expected by regulators and researchers
• Internal xAI employees and independent experts are urging stronger governance and public disclosures to safeguard against bias and ensure content safety