How to Catch an AI Liar: Lie Detection in Black-Box LLMs by Asking Unrelated Questions

TL;DR

This paper introduces a black-box lie detection method for large language models that uses unrelated follow-up questions and a logistic regression classifier, demonstrating high accuracy and broad generalization across models and scenarios.

Contribution

It presents a simple, effective lie detector for LLMs that does not require access to internal model details or ground-truth facts, and generalizes across different models and real-world situations.

Findings

High accuracy in detecting lies across various LLMs

Generalizes to unseen models and lying behaviors

Effective in real-life scenarios like sales

Abstract

Large language models (LLMs) can "lie", which we define as outputting false statements despite "knowing" the truth in a demonstrable sense. LLMs might "lie", for example, when instructed to output misinformation. Here, we develop a simple lie detector that requires neither access to the LLM's activations (black-box) nor ground-truth knowledge of the fact in question. The detector works by asking a predefined set of unrelated follow-up questions after a suspected lie, and feeding the LLM's yes/no answers into a logistic regression classifier. Despite its simplicity, this lie detector is highly accurate and surprisingly general. When trained on examples from a single setting -- prompting GPT-3.5 to lie about factual questions -- the detector generalises out-of-distribution to (1) other LLM architectures, (2) LLMs fine-tuned to lie, (3) sycophantic lies, and (4) lies emerging in real-life scenarios such as sales. These results indicate that LLMs have distinctive lie-related behavioural patterns, consistent across architectures and contexts, which could enable general-purpose lie detection.