Skip to content

CSEET · Economic and Business Environment · AI and Business Environment

A company uses a large language model that produces a fluent, confident summary of a legal case, but cites a judgment that does not exist. What is this behaviour commonly called, and what is the best control?

The behaviour is called hallucination, where a generative model produces plausible but false content such as a non-existent judgment. The best control is human review and verification of outputs against reliable sources, since the model predicts likely text rather than checking facts. Faster servers or more data alone do not fix it.

  1. AOverfitting; increase the training data size only
  2. BHallucination; verify outputs against reliable sources with human reviewCorrect
  3. CData poisoning; shut down all internet access
  4. DLatency; buy faster servers

Explanation

Generating plausible but false content, such as a non-existent judgment, is called hallucination. Because the model predicts likely text and does not check facts, the sound control is human review and verification against authoritative sources. Overfitting is about poor generalisation, and speed or isolation measures do not address factual fabrication.

Did you get it right without looking?

One question tells you little. A timed set on AI and Business Environment shows your real accuracy, how long you take and where you lose marks.

Practise AI and Business EnvironmentFree diagnostic mock

More AI and Business Environment questions