Research Summary: Cyber-Financial Contagion: Modeling the Propagation of an AI Vendor Compromise Through the Banking System

Research indicates that the banking system's increasing reliance on a limited number of shared artificial intelligence (AI) vendors introduces a significant systemic risk. A compromise within one of these vendors, affecting services like fraud screening, credit decisions, and anti-money laundering, could propagate through operational, informational, and financial linkages. This propagation could trigger cascading losses across financial institutions, potentially manifesting as a classical banking crisis, as demonstrated by a new stochastic epidemic-and-clearing model.

Why it matters

This research highlights a critical emerging systemic risk in the financial sector stemming from concentrated reliance on external AI vendors. Understanding these propagation mechanisms is crucial for developing robust risk management strategies and regulatory frameworks to prevent widespread financial instability triggered by cyber incidents or operational failures in third-party technology providers.

Key insights

• The banking system relies on a small number of shared AI vendors for critical functions.
• Compromise of an AI vendor can propagate through operational, informational, and financial linkages within the banking system.
• Such propagation can lead to widespread losses, potentially resembling a classical banking crisis.
• A four-layer heterogeneous network model (AI vendors, financial institutions, interbank exposures, customer accounts) has been developed to study this phenomenon.
• A stochastic epidemic-and-clearing model, CFC-Prop, simulates the propagation of compromise and its financial impact.

Source

arXiv — Computers and Society — https://arxiv.org/abs/2609.10350