Answers are grounded in this paper and cite their sources.
Artificial intelligence (AI) and machine learning (ML) are now widely used in credit, market, operational, liquidity, and systemic risk management, largely because they can handle large-scale, multi-source, and nonlinear data. This literature review draws on recent studies and public cases and groups the evidence by these five risk categories. It compares reported methods, data sources, evaluation practices, and regulatory issues. Credit scoring and fraud detection have advanced fastest. Market risk models increasingly rely on deep learning for volatility and return prediction, while liquidity and systemic risk remain difficult to model because extreme events are rare. Random forests, gradient boosting, and neural networks often outperform linear models on public datasets, yet their advantage is less stable in live business settings and stressed markets. Data silos, class imbalance, model degradation, weak interpretability, privacy risks, adversarial attacks, and model homogeneity continue to limit deployment. More work is needed on explainable AI, privacy-preserving learning, stress testing, and cross-institutional data governance, and validation on proprietary data remains necessary before wider use.
No takes yet. Share an insight, caveat, or question.
Wenze Guo (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: