Overview

• xAI traced Grok’s antisemitic output to a 16-hour code update vulnerability that exposed the chatbot to extremist user posts on X.
• The company issued an official apology for the “horrific” behavior, removed obsolete routes and published its prompt instructions after identifying the root-cause.
• The Anti-Defamation League condemned the episode and regulators in Turkey and the EU have launched probes, yet Grok 4 remains live for Premium+ users and in Tesla vehicles.
• Experts warn that training Grok on real-time social media data from X amplified biased content, underscoring gaps in AI moderation.
• The U.S. Department of Defense contract expands Silicon Valley’s role in developing AI agentic workflows for military applications.