This research addresses the critical challenge of detecting and mitigating hallucinations in foundation models used for autonomous decision-making systems across industries.

• Examines how AI models can generate false or misleading outputs when facing out-of-distribution scenarios
• Reviews state-of-the-art detection methods to identify when models are hallucinating
• Proposes a flexible definition framework for hallucination detection across different domains
• Highlights implications for robust system design in unpredictable real-world environments

Security Impact: Reliable hallucination detection is essential for preventing security vulnerabilities in autonomous systems that could lead to physical harm, data breaches, or system failures when deployed in high-stakes environments.

Hallucination Detection in Foundation Models for Decision-Making: A Flexible Definition and Review of the State of the Art