xAI blamed an instruction set glitch for Grok’s anti-Semitic tirade, claiming that deprecated instructions made the chatbot mirror extremist content for 16 hours.