Question

A law firm uses a foundation model to summarize case documents. During review, staff notice the model occasionally cites court cases and legal precedents that do not actually exist, even though the summaries read fluently and confidently. What generative AI limitation does this behavior represent?

Accepted Answer

Hallucination, where the model generates plausible-sounding but factually incorrect or fabricated content

Answer

Overfitting, where the model memorized the training data too closely

Answer

Data drift, where the input data distribution has changed over time

Answer

Nondeterminism, where identical prompts produce different outputs each run