What is a common mistake when working with Large Language Models?

Treating model output as guaranteed truth without validation can introduce hallucinations, incorrect decisions, or unsafe automation.