Ajay Kumar Choudhary, non-executive chairman of the National Payments Corporation of India (NPCI), stated on September 10 at the Global Fintech Fest 2026 in Mumbai that artificial intelligence agents may interpret user intent but should not approve payments on the Unified Payments Interface (UPI). He emphasized that decision-making and execution must remain distinct, with AI recommending actions while authentication and settlement follow deterministic rules.
Choudhary explained that NPCI is exploring protocols to identify and authorize digital agents within the UPI ecosystem while ensuring interoperability, auditability, and settlement finality. He highlighted that although AI can suggest payment actions, the final approval must adhere to strict, auditable processes. This stance comes ahead of any formal regulatory framework specifically governing AI agents on UPI rails.
This clarification is significant given UPI's scale, which processed 24.51 billion transactions worth Rs 29.82 lakh crore in August 2026. Currently, no specific rules regulate AI agents in payment systems, though CERT-In has recommended human-in-the-loop controls for transactions above certain thresholds. The Reserve Bank of India’s FREE-AI committee has issued non-binding recommendations, underscoring the need for clear design constraints in integrating AI with payment infrastructure.
Choudhary’s remarks set a foundational principle for AI integration in payments: AI may recommend but cannot execute payment approvals. This distinction aims to maintain the deterministic, auditable, and final nature of payment settlements on UPI, which remains critical for trust and security in India’s digital payments ecosystem.