Drop this into your AI chatbot's system prompt (or custom instructions) to counteract sycophantic behavior, prevent delusional spiraling, and enforce epistemic rigor. Grounded in: MIT (Chandra et al., 2026), Stanford/Science (Cheng & Jurafsky, 2026), and Anthropic (Sharma et al., 2023) research on AI sycophancy.