NPCI Launches AI Chatbot MyUPI for UPI Transactions
The National Payments Corporation of India (NPCI) has launched its AI-powered MyUPI chatbot to assist users with UPI transactions and mandates. Initial testing revealed some features are not yet functional, and data usage remains unclear.

The National Payments Corporation of India (NPCI) has introduced MyUPI, an AI-powered chatbot designed to support users of the Unified Payments Interface (UPI). Launched following its announcement at the Global Fintech Fest 2026, MyUPI aims to provide a consolidated view of UPI transactions and AutoPay mandates across various third-party apps.
The service is built on NPCI's proprietary "Small Language Model" named FiMI (Finance Model for India). NPCI states that MyUPI is intended as a support and service platform for end-users, offering assistance and insights. However, a disclaimer notes that any information provided is for support purposes only and should not substitute independent judgment.
Initial testing of the platform has highlighted several issues. Key announced features, such as the "Safety Switch" for declining transactions and "Automated Chargeback Processing," do not appear to be operational. Furthermore, the extent to which user transaction data is used for training NPCI's FiMI model remains ambiguous, despite recently updated privacy policies and terms of use.
The underlying technology of MyUPI incorporates various third-party models, including Mistral Small 24B Instruct and Llama 3, raising questions about its core dependencies. While NPCI promises 24/7 multilingual support, the chatbot currently supports only five Indian languages, and language switching affects the entire user interface.
MyUPI's privacy policy details the categories of data collected, including identity, transaction, device, and user interaction information. However, the policy does not explicitly clarify how this data is utilized for the training of the AI model, leaving a significant question regarding data privacy and AI development practices.