Study Highlights Persistent Hallucinations in Legal AI Systems
Description
Stanford University’s Human-Centered AI Institute (HAI) conducted a study in which they designed a "pre-registered dataset of over 200 open-ended legal queries" to test AI products by LexisNexis (creator of Lexis+ AI) and Thomson Reuters (creator of Westlaw AI-Assisted Research and Ask Practical Law AI). The researchers found that these legal models hallucinate in 1 out of 6 (or more) benchmarking queries.
Actors
Alleged Deployers
- legal-professionals
- law-firms
- organizations-requiring-legal-research
Alleged Developers
- thomson-reuters
- lexisnexis
Harmed Parties
- legal-professionals
- clients-of-lawyers
- legal-system
Classification
Risk Domain
MisinformationRisk Subdomain
3.1. False or misleading information
MIT Risk Profile
Source
Related Incidents (Misinformation)
View all →Research Papers on This Topic
Fast-DetectGPT: Efficient Zero-Shot Detection of Machine-Generated Text via Conditional Probability Curvature
Why Language Models Hallucinate
CIFAKE: Image Classification and Explainable Identification of AI-Generated Synthetic Images
Degenerate Feedback Loops in Recommender Systems