Glossary — Security

Hallucination

A hallucination is when an AI states something confidently that simply is not true — an invented source, a wrong figure, a law that does not exist.

3 tokensHallucination

Simply put#

A model does not distinguish truth from plausible-sounding text. It predicts what sounds right. Usually plausible is also correct — but not always, and the model does not know which just happened.

Example#

Ask for citations and you will get citations that look exactly like real ones: author, year, journal. Some of them may not exist.

Why it matters to you#

Always verify numbers, names, dates, legal provisions and references. A confident tone is not evidence.

Related terms