Banking Fraud: Are We Doing Enough to Prevent It, or Simply Getting Better at Detecting It?
Banks and financial institutions have invested significantly in fraud detection, cybersecurity, transaction monitoring and authentication. Banking today is more secure and technologically advanced in many respects than ever before.
Yet, hardly a day goes by without reports of customers losing money to fraud across UPI, credit cards, internet banking and digital wallets. Increasingly, fraudsters do not even need to compromise a bank's systems. A convincing phone call, fraudulent payment link or inadvertently shared OTP can be enough.
This raises a fundamental question. If banks already have sophisticated systems and controls, why are fraudsters continuing to find ways around them? More importantly, are we doing enough to prevent fraud, rather than investigating it after the damage has been done?
Beyond Detection and Investigation
According to the Reserve Bank of India's Annual Report 2024–25, banks and financial institutions reported 23,953 fraud cases involving ?36,014 crore. Card and internet fraud accounted for 56.5% of cases by number, while advances accounted for 92.1% of the amount involved.
These figures need context. Fraud reported in a particular year may have occurred earlier, and the amount involved does not necessarily represent actual financial losses. Nevertheless, the numbers underline the scale of the challenge.
Consider what happens when a customer reports fraud. Banks investigate, transaction trails are examined, cybercrime authorities may become involved, and attempts are made to recover the funds. All of this is necessary but often begins after the money has already moved through multiple accounts.
Having worked across banking operations, payments and business and technology transformation, I have seen how much effort organizations invest in strengthening individual systems and controls. The harder challenge is often making them work together effectively, particularly when decisions need to be made quickly.
I believe fraud management presents a similar challenge.
Are we measuring success by the number of suspicious transactions detected and cases investigated, or by how many fraudulent transactions were actually prevented?
Detection is a capability. Prevention is the outcome that ultimately matters.
Connecting Intelligence With Action
A fraudulent transaction may appear perfectly legitimate when viewed in isolation. A customer could be using a familiar device, authenticating successfully and approving a transfer, without realizing they are being manipulated.
But what if the beneficiary has received similar payments from several unrelated customers? What if the customer's behavior has changed significantly? Or what if account relationships and transaction patterns suggest a wider network of suspicious activity?
Banks already have considerable information across customer records, transactions, identity verification, cybersecurity and historical fraud investigations. The opportunity lies in connecting these signals to enable timely, informed decisions.
The Bank for International Settlements' Project Hertha demonstrated the potential of this approach. In experiments using synthetic data, payment-system analytics helped identify 12% more illicit accounts and improved identification of previously unseen financial crime patterns by 26%. While experimental, the findings highlight the potential of connected intelligence.
This is where AI, behavioral analytics and real-time decision-making can make a difference. But simply deploying more sophisticated AI models will not solve the problem if responses still depend on disconnected systems, manual investigations or delayed interventions.
The real opportunity is to connect intelligence with the ability to act before a loss occurs.
That could mean additional verification, temporarily holding a high-risk transaction where permitted, or warning a customer when behavioral patterns suggest possible manipulation all without unnecessarily inconveniencing legitimate customers.
Protecting Customers Across the Ecosystem
Fraud prevention also raises important questions about customer protection and accountability.
A fraudulent transaction may originate through a third-party application or result from a customer being persuaded to authorize a payment. The bank's systems may have functioned as designed, yet the customer still suffers a loss.
Under RBI's customer-protection framework, liability for unauthorized electronic banking transactions depends on the circumstances, including customer negligence and how promptly the incident is reported.
But from a customer's perspective, the distinction between a compromised banking system and a sophisticated social engineering attack offers little comfort.
If a customer has been manipulated into authorizing a payment, is a successfully entered OTP sufficient assurance that the transaction is genuinely safe?
This is why prevention must extend beyond individual banks, with stronger collaboration among payment networks, fintechs, merchants and regulators to identify suspicious accounts and emerging fraud patterns.
What Needs to Change?
I see three priorities.
First, build prevention into the customer journey. Stronger identity protection, secure architecture, intelligent authentication and continuous risk assessment must work together rather than operate as isolated controls.
Second, shorten the time between identifying a risk and taking action. AI and analytics can improve detection, but their real value comes when insights translate into timely intervention.
Third, address fraud as an ecosystem and operating-model challenge. Fraud management, cybersecurity, AML, payments and customer experience need greater integration, shared intelligence and clearer decision-making responsibilities.
Technology and transformation partners can contribute through capabilities across data engineering, AI and analytics, cybersecurity, systems integration and technology modernization, helping institutions connect existing investments and strengthen preventive controls.
Ultimately, the success of these investments should be measured through fewer successful fraudulent transactions, lower losses, faster intervention and stronger customer protection.
The Real Measure of Progress
Financial fraud will continue to evolve, particularly with AI-enabled impersonation, deepfakes and increasingly sophisticated social engineering. In April 2025, Federal Reserve Vice Chair for Supervision Michael Barr cited estimates of a twenty-fold increase in deepfake attacks over three years.
Eliminating fraud altogether may be unrealistic. But with the advances in AI, data and real-time payments, we should certainly be aiming to make fraud harder to perpetrate and easier to interrupt.
Perhaps the question we should be asking is not whether banks have invested enough in fraud detection, but whether those investments are doing enough to prevent customers from losing their money.
Because ultimately, that is the outcome that matters.
Author Bio
Anil Khabiya leads Sales and Business Development for Solutions at Dexian India. With over 24 years of experience, predominantly in Banking and Financial Services, he brings together deep banking domain expertise, management consulting experience and business leadership.
His career spans HDFC Bank, HSBC and ABN AMRO, followed by consulting and technology leadership roles at Accenture and Expleo. His experience includes retail banking, payments, banking operations, M&A integration, business and technology transformation, and operating-model design across India, the UK, Europe, the Middle East and APAC. He focuses on bridging business priorities and technology-led transformation to deliver measurable outcomes.