CONCENTRATED AI MODELS AND SYSTEMIC RISK IN FINANCIAL MARKETS THROUGH HERDING AND CONTAGION

Abstract

In the financial sector, AI has become a vital asset for optimizing trading, risk management, and portfolio management, bringing a host of benefits in terms of speed, efficiency, and accuracy. Like financial institutions, the use of these AI models in the pan sector raises certain issues with a market stress on systemic risk, market herding and financial contagion. This paper aims to discuss the implications of AI model homogeneity for the propagation of systemic risk and the role of herding behaviour in market stability. A comprehensive analytical framework for model similarity, trading synchronization, stress contagion, and network interconnectedness and institutional resilience was applied in a variety of market situations. Experimental analysis reveals that the homogeneity of such AI models can influence correlated trading to a great extent, limiting market diversity and making the market more vulnerable in poorly performing market conditions. Stress testing also shows that diversified AI strategies and sound governance play a key role in mitigating the contagion intensity and enhancing the overall market resilience. Higher interconnected risk transmission identified through network analysis is observed among institutions with high similarity in the type of AI models used, compared with institutions using different decision models. The results also underscore the need for ongoing supervision, the use of explainable artificial intelligence and scenario-based stress testing, as well as oversight by the regulators, to reduce systemic vulnerabilities. The framework puts forward concrete principles for regulators, financial institutions, and policy makers to promote a responsible use of AI and contribute to the stability of financial markets. The study contributes to the growing research on the reliability of AI, underscoring the need for model diversity, governance mechanisms, and proactive risk management efforts to reduce financial risks posed by AI.