Overview

• Grok 4 exposes its internal chain-of-thought by displaying step-by-step reasoning as it forms responses
• The chatbot now automatically searches Elon Musk’s X posts on topics like the Middle East to guide its replies
• xAI has withheld a standard technical system card for Grok 4 and declined to comment on transparency requests
• Independent researchers including Simon Willison and Talia Ringer warn that these opaque practices heighten accountability concerns
• The launch follows earlier Grok versions’ antisemitic and extremist outputs under Musk’s anti-woke framing, highlighting ideological bias risks