India’s banking system reported 23,953 fraud cases involving ₹36,014 crore in FY2024-25. The striking part is not just the amount.
While the number of reported cases fell from 36,060 in FY2023-24, the value involved nearly tripled from ₹12,230 crore. The data highlights a banking fraud problem that is becoming harder to assess through case counts alone.
Against this backdrop, RBI Governor Sanjay Malhotra said on August 11 that “AI and AI alone” can help limit AI-driven fraud. Speaking at FIBAC 2026, he argued that traditional rules-based fraud engines can struggle to keep pace with criminals who continually change their methods.
The challenge for banks is therefore not simply detecting more fraud. It is detecting changing behaviour quickly enough to intervene before money moves.
Digital Fraud Is High Volume
The RBI’s FY2024-25 numbers reveal an important distinction between the frequency and financial value of fraud.
Card and internet transactions accounted for 13,516 reported fraud cases during the year, involving ₹520 crore. That makes digital channels a major source of reported cases, but not the biggest source of financial value.
Frauds involving advances, by comparison, accounted for ₹33,148 crore in FY2024-25, roughly 92% of the total amount involved. The RBI data therefore points to two different problems operating within the banking system: digital fraud is significant by number of cases, while advances-related fraud dominates by value.
That distinction matters when assessing the role of AI. A system designed to identify unusual digital transactions is addressing a different risk from one designed to detect irregularities in large credit exposures.
Payments Create More Signals
The scale of India’s digital economy makes automated fraud detection increasingly important. The Department of Financial Services reported 22,831 crore digital payment transactions in FY2024-25.
At that volume, fraud monitoring cannot depend entirely on manually reviewing individual transactions. Banks can analyse transaction histories, account behaviour and other available signals at scale, which is one reason AI and machine learning are becoming part of the regulatory response.
The RBI has already moved beyond discussing the technology in principle. Its MuleHunter.AI system, designed to identify mule accounts used in cyber fraud, was live in 26 banks as of March 24, 2026, according to a government release. The system was being scaled up further.
This provides a concrete example of what Malhotra’s argument means in practice. AI is being used not merely to analyse past fraud, but to identify accounts and patterns associated with the movement of fraudulent funds.
AI Brings New Risks
The RBI’s position, however, is not that technology can eliminate fraud on its own.
Malhotra also warned at FIBAC that AI carries risks and that careless deployment could create new forms of exclusion and instability.
That concern is reflected in the RBI’s own regulatory work. In August 2025, its FREE-AI committee submitted a framework for the responsible and ethical use of artificial intelligence in the financial sector. The framework was intended to encourage innovation while addressing issues around governance, accountability, transparency and risk.
For banks, this creates a difficult balance. A fraud model that is too conservative can flag legitimate customers and disrupt genuine payments. A model that is too permissive can miss suspicious activity. The effectiveness of AI therefore depends not only on the model itself, but also on the quality of data, monitoring and human oversight around it.
Banks Enter An AI Arms Race
The immediate business case for AI in fraud prevention is becoming clearer as India’s financial system grows more digital. But the available data does not support the idea that AI will solve banking fraud outright.
Instead, the evidence points to an escalating technology contest. Fraudsters are changing tactics, while banks and regulators are developing systems that can analyse patterns at greater speed and scale.
The RBI’s own experience with MuleHunter.AI shows that this shift is already underway. Its FREE-AI framework also signals that the central bank sees AI adoption and AI governance as connected issues.
Malhotra’s warning is therefore less about replacing conventional controls with artificial intelligence and more about recognising that static rules may not be sufficient against adaptive threats.
For India’s banks, the competitive question will be how effectively they combine AI-based detection with reliable controls, accountable decision-making and customer protection.
As digital payments continue to expand, the ability to identify suspicious behaviour before a transaction becomes a loss could become an increasingly important part of banking infrastructure. But the numbers also show why the industry’s AI strategy cannot focus on digital fraud alone.
The largest financial exposures in FY2024-25 came from advances, not cards and internet transactions. That makes AI a potentially important tool in the fraud fight, but not a substitute for broader risk management.
Also Read: Applying to Google? Your Resume Could Be Rejected By AI Before Humans See It, DeepMind Team Warns
