FinRisk AgentNet: Multi-Agent LLM & ML Architecture for Agentic Fraud Risk Detection & Quantitative Financial Risk Management

Abstract

The rapid digitization of financial ecosystems has significantly increased the scale, speed, and complexity of fraud and risk events, making conventional rule-based monitoring and siloed machine learning systems insufficient for modern financial institutions. This study proposes FinRisk AgentNet, a unified multi-agent architecture that integrates Large Language Models (LLMs) with machine learning (ML) techniques for agentic fraud risk detection and quantitative financial risk management. The framework is designed to operate across heterogeneous financial data streams, including transactional records, customer behavior logs, market indicators, compliance reports, and unstructured textual evidence such as case notes and alerts. FinRisk AgentNet employs a coordinated society of specialized agents, including fraud detection agents, anomaly scoring agents, risk forecasting agents, compliance reasoning agents, and decision orchestration agents, each responsible for a specific analytical task while collaborating through a shared memory and policy layer. The proposed architecture combines supervised fraud classification, unsupervised anomaly detection, graph-based relationship analysis, and time-series risk forecasting with LLM-driven reasoning, contextual interpretation, and alert summarization. This hybrid design enables the system not only to detect suspicious transactions and emerging fraud patterns in real time but also to quantify broader financial risks such as credit risk, liquidity exposure, and operational risk. A key contribution of the model lies in its ability to fuse structured numerical risk signals with unstructured semantic evidence, thereby improving decision quality, explainability, and response speed. FinRisk AgentNet also introduces an adaptive feedback loop in which agent outputs are continuously evaluated and refined using risk thresholds, confidence scores, and historical outcomes. The framework supports human-in-the-loop governance, regulatory traceability, and scalable deployment in banking, insurance, and fintech environments.

REFERENCES

1. Rahman, M., Bhuiyan, A., Islam, M. S., Laskar, M. T. R., Mahbub, R., Masry, A., ... & Hoque, E. (2025). Llm-based data science agents: A survey of capabilities, challenges, and future directions. arXiv preprint arXiv:2510.04023.
2. Spears, T., Hansen, K. B., Xu, R., & Millo, Y. (2025). Governing synthetic data in the financial sector. Finance and Society, 1-17.
3. Joshi, S. (2025). Comprehensive review of artificial general intelligence (AGI): Applications in business and finance.
4. Karras, A., Theodorakopoulos, L., Karras, C., Krimpas, G. A., Giannaros, A., & Bakalis, C. P. (2025). LLM-Driven Big Data Management Across Digital Governance, Marketing, and Accounting: A Spark-Orchestrated Framework. Algorithms, 18(12), 791.
5. Lopez, E. (2024). Towards eXplainable Artificial Intelligence (XAI) in cybersecurity(Doctoral dissertation).
6. Ribeiro, R. N. (2025). Exploring the Feasibility of Using Large Language Models for Dark Web Threat Intelligence in Security and Defence(Master's thesis, Universidade do Porto (Portugal)).
7. Wang, Q., Wang, T., Tang, Z., Li, Q., Chen, N., Liang, J., & He, B. (2025, July). MegaAgent: A large-scale autonomous LLM-based multi-agent system without predefined SOPs. In Findings of the Association for Computational Linguistics: ACL 2025(pp. 4998-5036).
8. de la Roche, M., Voloder, E., Banerjee, A., Guerra, C., Cataldo Dell'Accio, D., Budris, F., ... & Sedej, T. (2024). Report on artificial intelligence and blockchain convergences. Available at SSRN 5023415.
9. Li, H., Huang, S., & Park, J. K. (2025). A Human-Centered Review of Large Language Models for Online Scam Detection. Available at SSRN 6338940.
10. Steindl-Haselbauer, K. (2025). Potentials, risks and implications of AI-driven Strategic Foresight in disruptive environments-A framework for companies to exploit new opportunities and remain resilient(Doctoral dissertation, Technische Universität Wien).
11. Adabara, I., Sadiq, B. O., Shuaibu, A. N., Danjuma, Y. I., & Venkateswarlu, M. (2025). A Review of Agentic AI in Cybersecurity: Cognitive Autonomy, Ethical Governance, and Quantum-Resilient Defense. F1000Research, 14, 843.
12. Lei, M., Liu, Q., Li, K., Zhao, J., & Tian, Z. THE STUDY OF FEASIBILITY AND CHALLENGES ON SILICON-BASED LIFEFORMS BASED ON LARGE LANGUAGE MODEL.
13. Raza, S., Qureshi, R., Zahid, A., Muneer, A., Zafar, A., Kamawal, S., ... & Shoman, M. (2025). Who is responsible? the data, models, users or regulations? a comprehensive survey on responsible generative ai for a sustainable future. arXiv preprint arXiv:2502.08650.
14. Pamisetty, V. (2023). Leveraging AI, Big Data, and Cloud Computing for Enhanced Tax Compliance, Fraud Detection, and Fiscal Impact Analysis in Government Financial Management. Fraud Detection, and Fiscal Impact Analysis in Government Financial Management (December 15, 2023).
15. Nastoska, A., Jancheska, B., Rizinski, M., & Trajanov, D. (2025). Evaluating trustworthiness in AI: Risks, metrics, and applications across industries. Electronics, 14(13), 2717.
16. Kolla, S. H. (2023). Deep Learning–Driven Retrieval-Augmented Generation for Enterprise ITSM Automation: A Governance-Aligned Large Language Model Architecture. Journal of Computational Analysis and Applications, 31(4).
17. Abdallah, A. A., Aslan, H. K., Abdallah, M. S., Cho, Y. I., & Azer, M. A. (2025). Enhancing Cryptocurrency Security: Leveraging Embeddings and Large Language Models for Creating Cryptocurrency Security Expert Systems. Symmetry, 17(4), 496.
18. Motamary, S. (2024). Transforming customer experience in telecom: Agentic AI-driven BSS solutions for hyper-personalized service delivery. Available at SSRN 5240126.
19. Vlachos, I., & Reddy, P. G. (2025). Machine learning in supply chain management: systematic literature review and future research agenda. International Journal of Production Research, 63(16), 5987-6016.
20. Moussaoui, J. E., Kmiti, M., El Gholami, K., & Maleh, Y. (2025). A systematic review on hybrid AI models integrating machine learning and federated learning. Journal of Cybersecurity and Privacy, 5(3), 41.
21. Wicker, M., Szpruch, L., & Mørk, S. (2025). Move Fast without Breaking the Bank Model Risk Management of GenAI workflows.
22. IKE, P. C., DANIEL, U. C., ADEOYE, M., AMAECHI, M. I., ODESOLA, O. O., & OGUNSAKIN, M. O. (2025). COGNITIVE CYBER DEFENSE SYSTEMS: LEVERAGING ARTIFICIAL INTELLIGENCE FOR AUTONOMOUS THREAT PREDICTION AND RESPONSE. International Journal of Nature and Science Advance Research.