India’s digital payments system is getting an AI upgrade just as banks face tougher fraud rules.
Dilip Asbe, CEO of the National Payments Corporation of India, told TechCrunch that AI will play a major role in the next phase of UPI growth, from onboarding new users to finding fraud and mule accounts. UPI, India’s real-time payments system, now handles more than 750 million daily transactions, with a goal of passing 1 billion daily transactions.
India is also tightening protections for victims of digital fraud. New Reserve Bank of India rules, reported by the Economic Times, take effect on January 1, 2027, and set new expectations for compensation, complaint handling, and bank response times.
NPCI looks to smaller AI models
Asbe said India’s payments industry has an opportunity to build smaller, more focused language models rather than chasing giant general-purpose systems. He told TechCrunch that banks, fintech companies, and ecosystem players can build models that are “sharp, specific, and as deterministic as possible.”
That kind of focus is important in payments. A customer dispute tool cannot wander through open-ended answers. It needs to identify the right workflow, surface the correct information, and help resolve the issue quickly.
NPCI has already moved in that direction with FiMI, a finance-focused model used for payment support. TechCrunch reported that FiMI is serving more than 1 million users for tasks such as canceling mandates and resolving issues.
The approach fits a broader enterprise AI shift. Companies are finding that AI is often most useful when it is tied to a narrow operational process, not dropped into every workflow at once. In finance, AI trading agents are already pushing automation into money movement, though payments infrastructure carries a different level of operational risk.
Fraud rules raise the stakes
The RBI’s new framework adds pressure on banks to handle fraud complaints faster. The Economic Times reported that complaints must be investigated and resolved within 45 calendar days for domestic fraud transactions and 60 calendar days for cross-border cases.
The rules also create a compensation mechanism for smaller fraud losses. Victims who lose up to ₹50,000 ($529) may be eligible to receive up to ₹25,000 ($264) from banks once in their lifetime, according to the Economic Times.
Fraud detection is already becoming a major AI use case. The Times of India reported that India’s finance ministry directed banks to adopt MuleHunter.AI, an RBI-developed tool for identifying mule accounts used in cyber fraud.
For banks and payment operators, the message is straightforward: AI can help scale customer support and fraud detection, but the systems need to be narrow, auditable, and tied to clear rules. Warnings about an AI-powered fraud crisis in banking make that balance harder to ignore.
India’s payments stack is becoming a test case for regulated AI in financial operations. NPCI is trying to scale AI inside one of the world’s busiest real-time payment systems, while the RBI is raising the cost of slow fraud response. The country is also drawing more AI infrastructure attention, including OpenAI’s reported 100 MW India deal.
The result could shape how other payment networks use AI for support, fraud screening, and dispute handling.
Also read: Gemini in Chrome across APAC brings Google’s browser AI tools to more users in India, Indonesia, Malaysia, the Philippines, Singapore, and Vietnam.