Overview
- RBI’s FREE-AI framework and SEBI guidance are cited as speeding adoption while requiring explainable models, data governance and independent oversight.
- The report says more than 15 banks use the RBI Innovation Hub’s MuleHunter tool, with one bank reporting about 95% accuracy in identifying mule accounts.
- Banks are replacing threshold rules that trigger high false positives with machine learning for real-time monitoring, anomaly detection, behavioral analytics and automated SAR drafting using NLP.
- KPMG warns that regulators are intensifying model risk management expectations, urging independent validation and stress-testing to prevent AI from introducing systemic risks.
- An academic study by Chuanhao Nie, Yunbo Liu and Chao Wang reports a Graph RAG agent beat a standard AI on complex multi-hop AML reasoning in a synthetic banking test, while noting privacy and explainability remain hurdles.