Hallucination (making things up)
Plausible, confident and wrong, most often in specific details.
A model fills gaps with plausible text. It shows up most in specifics: statistics, quotes, names, dates, links and sources. Lawyers have been sanctioned for citing court cases an AI invented.
The danger is not that it is sometimes wrong. It is that the tone never changes. The output is equally fluent whether it is drawn from ten thousand pages or from nothing. Three habits catch most of it: supply the source text, ask for the exact supporting quote, then spot-check two or three by hand.
Find your way around
Learn it in the course, copy it from a playbook, see it work in a case study.