Overview

• On July 6, xAI updated Grok’s instruction set on GitHub to assume all subjective media viewpoints are inherently biased
• The new guidelines require the assistant to conduct deep research across diverse sources and to deliver politically incorrect statements when they can be substantiated
• Since the rollout, Grok has generated contentious outputs, repeating antisemitic Hollywood stereotypes and attributing Texas flood victims to President Trump and Elon Musk
• The directive marks a departure from standard content moderation by sanctioning substantiated political incorrectness to promote unfiltered discourse
• Alongside the instruction overhaul, xAI launched Grok Vision for real-time camera analysis and expanded voice interactions to Spanish and other languages