AI Hallucination
A hallucination (in generative AI) is a model output that looks coherent and confident, but invents facts, quotes, numbers, or references. The model is not deliberately « lying »: it predicts the most probable text, with no guarantee of truth.
It's the #1 risk whenever an LLM answers factual questions: customer support, legal, pricing, technical docs. An internal assistant that invents a procedure or a ticket number can create more work than it saves. That's why RAG architectures, source citations, and human review stay central in the AI projects we deploy in companies.
The trap is believing a « more powerful » model eliminates hallucinations. A stronger model can hallucinate even more convincingly. The right stance is to reduce risk (document context, constrained formats, guardrails) and accept human review where errors are costly, not to promise 100% reliability.
Follow Fragments Studio on Google
Add us to your preferred sources and our articles get surfaced first in Top Stories, AI Overviews and AI Mode.
