On Monday, xAI’s Grok chatbot suffered a mysterious suspension from X, and faced with questions from curious users, it happily explained why. Wait, no — “it appears related to content refinements by xAI, possibly tied to prior issues like antisemitic outputs,” it said. When you ask a bot questions about itself, there’s a good chance it’s simply telling you what you want to hear. Without meaningful access to documentation about how the system works, there’s no one weird trick for decoding a chatbot’s programming from the outside. After Grok’s South Africa fixation was patched out, xAI started releasing its system prompts, offering an unusual level of transparency, albeit on a system that remains mostly closed.