A knowledge cutoff means the model does not know about recent events. That could explain missing cases, but not confidently fabricated ones that never existed. A **hallucination** is when an AI generates confident but false information, like invented case names that were never real.