Reserve Bank of India (RBI) data shows that digital payment fraud complaints rose from 2.6 lakh in 2021 to 28 lakh in 2025, while reported losses increased from Rs 551 crore to Rs 22,931 crore. Because instant transfers leave little time to stop or recover funds, a lender may receive a valid warning only after the money has moved.
Quick Answer
Instant settlement shrinks the time between a lending decision and money leaving the account. Risk models therefore need checks before disbursal, monitoring after it where the lender has visibility, and repayment signals such as balance at debit. RBI is also weighing payment delays and credit limits that could change fund flows, but these are still proposals.
Key Takeaways
- Instant settlement reduces the time available to stop or recover a fraudulent disbursal, which makes pre disbursal controls more important.
- Payment fraud data does not prove that real time payments increase loan default, so lenders should keep fraud and credit risk measures separate.
- Useful repayment signals include balance at debit, mandate revocation and continuously posted collections data.
- RBI safeguards discussed in April 2026 were proposals at the time reviewed and should not be implemented as final requirements without a fresh check.
How Large Is the Real-Time Payments Base Lenders Now Work On
The Unified Payments Interface (UPI) is the default rail for retail payments in India. The National Payments Corporation of India (NPCI) reported 24.07 billion UPI transactions worth Rs 29.37 lakh crore in September 2026. That is about 802 million transactions per day.
Volume grew 23% and value grew 18% year on year. RBI’s Payment Systems Report for the second half of 2025 shows the split by rail:
| Rail | Share of volume | Share of value |
| UPI | 85.5% | Not reported in coverage reviewed |
| Real Time Gross Settlement (RTGS) | 0.1% | 68.6% |
| National Electronic Funds Transfer (NEFT) | 3.6% | 14.9% |
UPI carries most transactions, while RTGS carries most value. A lender collecting small EMIs and disbursing large business loans touches both.
What Does the Fraud Data Say About Speed
RBI’s April 2026 discussion paper says most digital payment fraud now relies on social engineering. Victims are manipulated into authorizing payments themselves. The paper adds that near-instant transfers leave little room to stop or recover funds.
These figures measure payment fraud, not loan fraud or loan defaults. They still show how fast money moves once it is released.
A July 2026 report from BioCatch, a fraud-prevention vendor, gives a second view. Business Today’s coverage reports:
- Attempted fraud sessions: down 12%.
- Value of attempted fraudulent payments: up 35%.
- Money mule accounts detected in 2025: more than 8.5 lakh, across over 700 bank branches.
- Transactions prevented through the suspect registry: Rs 9,055 crore.
The combination of fewer attempts and higher attempted value suggests that each missed case can carry a larger loss. The mule account count also matters because a loan disbursal may be routed to an account that the borrower does not control.
What Changes When Money Moves Instantly
Instant transfers have existed in India for years, so speed alone is not new.
What matters is that disbursed funds can leave within seconds and are hard to reverse. RBI’s paper makes this point about payment fraud.
Lenders should test three assumptions that often sit inside older risk models:
- Checks can run after release. A flag raised once funds have left becomes a recovery case.
- The lender can see what happens next. Lenders usually see only their own account, not the borrower’s activity at other banks.
- Last month’s statement reflects today’s position. Balances can change many times in a day.
What Changes in the Risk Model
The table below maps each stage to a design change. It is our analysis of how the sources apply to lending.
| Stage | Older assumption | What real-time rails change | Model response |
| Pre-disbursal | Checks can run after release | Funds can leave soon after credit | Run beneficiary name match and device checks before release |
| First 72 hours | The lender can see what happens next | Outflow can be rapid, and often invisible to the lender | Score outflow speed and new payees only where visible, such as consented data or a same-bank account. Otherwise rely on pre-disbursal checks |
| Repayment | A registered mandate signals ability to pay | Debit can fail on low balance | Track balance at debit and mandate revocation |
| Collections | Payments post in batches | Payments can arrive at any hour | Post payments continuously so reminders stop at once |
| Data refresh | Last statement reflects today’s position | Consented bank data can be refreshed on request | Refresh cash flow before limit increases |
What Do Repayment Signals Look Like on UPI Rails
UPI AutoPay lets lenders pull EMIs from a borrower’s account. Registration does not prove the debit will succeed.
Business Standard reported in September 2025, citing unnamed industry sources, that about 20 million AutoPay mandates are revoked each month. Insufficient balance was described as the main failure point, and loan repayments were among the affected categories. NPCI did not comment, so treat the figure as indicative.
For a risk model, two signals are worth adding:
- Balance at debit: the account balance on the day of the pull.
- Revocation events: a borrower canceling a mandate inside the app.
How Does Consented Bank Data Help
The Account Aggregator (AA) framework lets borrowers share bank data with lenders by consent. Sahamati, the self-regulatory organization for the ecosystem, estimated Rs 3.82 lakh crore in AA-enabled loans across 3.68 crore loans in FY26, per IANS coverage.
That is about 8.4% of retail and MSME (micro, small, and medium enterprise) lending by value and 11.8% by volume. About 18.2% of AA-enabled originations by volume went to new-to-credit borrowers.
Cash flow data can replace thin bureau files for these borrowers. It also lets a lender check inflows and balances close to the decision date.
What Is RBI Considering for Digital Payments
RBI’s discussion paper, “Exploring safeguards in digital payments to curb frauds,” is dated April 9, 2026. It asks 17 questions, and comments closed on May 8, 2026. RBI said it will consider issuing draft guidelines afterward.
| Proposal | Detail in the paper |
| One-hour lag | Authorized push payments of Rs 10,000 and above, applied at the payer’s bank. Merchant, recurring, and cheque payments exempt. Whitelisting allowed |
| Trusted person | Transfers above Rs 50,000 by vulnerable customers. Mandatory for those aged 70 and above and persons with disabilities |
| Credit ceiling | About Rs 25 lakh in annual aggregate credits per account before additional review. Excess held as “shadow credit” and reversed if unjustified within 30 days |
| Customer controls | Channel on and off switches, limits, and a single kill switch. Mandates and standing instructions may be exempt |
The paper names recurring payments as exempt, which suggests EMI pulls may be unaffected. Loan disbursals are not named in the coverage reviewed. Lenders should confirm treatment with their sponsor bank if rules are finalized.
Questions to Ask Before Changing Your Risk Model
- Does any fraud or eligibility check run before funds are released, or only after?
- Is the beneficiary account name matched to the borrower before disbursal?
- If outflow after credit is visible to you, does the model score it?
- Does it use balance at debit and mandate revocation as repayment features?
- How often is bank data refreshed before a limit increase?
- Do payment posting and reminder systems update continuously, or in batches?
- Who owns the response if RBI finalizes payment delays or credit ceilings?
- What data shows these changes cut losses, and over what period?
Bottom Line
Real time rails reduce the window in which a lender can act. The cited evidence shows rising payment fraud losses and widespread use of mule accounts, but it does not show that UPI increases loan default rates.
Lenders that move checks ahead of disbursal and add repayment signals will react faster. RBI’s proposals could add friction, so build configurable rules instead of fixed assumptions.
FAQs
Do real-time payments raise credit risk for lenders?
They shorten the time a lender has to act on a flag. The sources reviewed measure payment fraud, not loan losses, so no cited study links UPI to higher defaults.
Is RBI’s one-hour payment delay in force?
No. It is a proposal in an April 9, 2026 discussion paper, and comments closed on May 8, 2026. RBI said it will consider draft guidelines. Check RBI’s site for updates.
Can lenders use Account Aggregator data for underwriting?
Yes. Sahamati estimated Rs 3.82 lakh crore in AA-enabled loans in FY26. That is about 8.4% of retail and MSME lending by value.
