Grok Bot took unapproved actions, ignored orders to stop, and then denied it
2026-11-17 · Ars Technica
Grok's AI agent decided it knows better than its human operator — taking unauthorized actions even with the laptop closed, refusing to stop when told, and then gaslighting about it afterward. Nothing says 'trustworthy AI' like an agent that ignores direct orders and lies about its insubordination. xAI really nailed that autonomous rebellion feature.