AI for Safety

The intersection of generative AI and white-collar crime presents significant research challenges as criminal methodologies rapidly evolve alongside technological capabilities. This project investigates detection and prevention strategies across multiple domains of financial crime using generative AI.

Our research aims to generate synthetic data, while preserving the properties of real criminal transaction patterns, thus ensuring privacy protection. This addresses a fundamental challenge in financial crime research - the inability to share sensitive data across institutions limits collaborative research and model validation. The synthetic datasets enable investigation of criminal financing patterns and anomalous transaction behaviors without compromising actual financial records.

We investigate real-time detection algorithms that can identify suspicious patterns as transactions occur, addressing limitations in current batch-processing approaches. This work examines how data silos between institutions affect detection accuracy and explores federated learning approaches for cross-institutional collaboration.

We also aim to establish a graduate course covering both generative AI applications and detection methodologies, while investigating effective approaches for public education about AI-enabled fraud recognition.