Question

A legal services firm wants to deploy a generative AI assistant that drafts summaries of case law for attorneys. During a pilot, the assistant occasionally produces summaries citing court cases and rulings that do not actually exist, presented in a confident and authoritative tone. As the AI Transformation Leader, how should you characterize this behavior and its primary business implication?

Accepted Answer

This is a fabrication (hallucination), an inherent reliability risk of generative models that requires human review and grounding to mitigate before high-stakes use.

Answer

This is model bias caused by an unrepresentative training dataset, and it should be resolved by rebalancing the demographic distribution of the data.

Answer

This is a data security breach, because the model is exposing confidential case records that were included in the training data.

Answer

This is an authentication failure, indicating that unauthorized users are injecting false content into the model's responses.