AI Hallucinations

An AI hallucination is when a language model produces confident but false or made-up information: an invented policy clause, a wrong leave balance, or a citation to a document that does not actually exist. It happens because the model generates whatever is statistically plausible rather than what is actually verified, so when it does not have the real answer, it can still produce a fluent, wrong one.

Why this matters more in HR

This is the number one risk for an enterprise assistant. A confident wrong answer about leave, pay, or policy is worse than simply saying "I don't know."

The fix is architectural, not a matter of the model trying harder: grounding the answer only in approved sources and saying so when those sources do not cover the question, guardrails, a low temperature setting for factual use cases, and a maker-checker design where a second model validates the first model's output against rules and thresholds before it ever reaches the user.