FraudSociety: A Generative Agent-Based Society for the Financial-Fraud Economy

Researchers have introduced FraudSociety, the first generative LLM-agent-based simulator designed to jointly model informational and financial dynamics within the financial-fraud economy, addressing empirical gaps that contributed to over $15.9 billion in reported US losses in 2025.

This development provides organizations and regulatory bodies with an integrated modeling tool to evaluate systemic fraud exposure before interventions are deployed. By simultaneously mapping communication channels and financial ledgers, decision-makers can better anticipate aggregate harm and test counter-fraud measures. It shifts fraud risk management from reactive tracking to proactive behavioral simulation.

Financial fraud caused more than $15.9 billion in losses in the United States in 2025 according to FTC data.

Forward consideration, not a verified fact.