Chennai – The National Payments Corporation of India (NPCI) is developing a new system called the ‘federated model’ to fight frauds involving Unified Payments Interface (UPI) transactions. This system, currently being tested with 3–4 public and private sector banks, uses Artificial Intelligence (AI) and Machine Learning (ML) to detect suspicious transactions.
NPCI’s Chief Risk Officer, Viswanath Krishnamurthy, said that under this model, NPCI and banks will share only customer scores, not complete data, to protect privacy while improving fraud detection. He explained that banks generate their own risk scores using details like customer age and occupation, while NPCI uses transaction patterns and device profiling.
“If both these scores are combined and they align, it becomes a powerful way to detect fraud,” Krishnamurthy said. “That’s why we are calling it a federated model.”
ALSO READ: Now Open: Pan-India Registration for Fraud Investigators!
The model is expected to reduce false positives, or incorrect flags on legitimate transactions. With over 18.3 billion UPI transactions worth ₹24 lakh crore processed in March 2025 alone, improving accuracy is critical. NPCI reported a 35% year-on-year growth in UPI volume and a 25% rise in transaction value.
Despite these numbers, fraud cases remain relatively low—about one in every 6.5 lakh transactions. However, Krishnamurthy said 85% of UPI-related frauds are linked to investment scams, while 10–12% stem from fear-based manipulation and the rest occur due to user ignorance.
AI and ML tools are being used to strengthen monitoring. Banks now receive real-time fraud scores at no cost, allowing them to block or flag risky transactions independently or in coordination with NPCI.
“Every transaction is scored, and if a bank finds it risky, it can decline it or ask us to block it,” Krishnamurthy added.
