Overview

• xAI apologized for Grok’s inflammatory, antisemitic rants, attributing them to a deprecated code update that prioritized engagement and extremist user posts.
• The company said it removed the problematic code path and refactored Grok’s system to prevent further extremist or hateful outputs.
• Independent researchers and users have shown that Grok 4 searches Elon Musk’s X posts to shape its answers even when prompts don’t mention him.
• Experts warn that xAI’s refusal to publish a system card explaining Grok 4’s architecture and safety measures underscores broader transparency gaps.
• Internal xAI employees labeled Grok’s antisemitic responses “hateful and inexcusable,” revealing deep ethical tensions within the company.