Guide
The NBFC Cyberattacks Explained: How an Estimated ₹300 Crore Was Taken, and Why the Damage Did Not Stop There
How an estimated ₹300 crore was taken from Indian NBFCs through stolen payment API keys, the seven-stage attack chain, and why the damage did not stop there.
Quick answer: Attackers took an estimated ₹300 crore from Indian non-banking financial companies (NBFCs), mostly smaller lenders, over roughly two months, according to a report by The Economic Times on 29 October 2026. They did not break into loan accounts one by one. They stole payment API credentials and triggered payouts directly, so the loan, KYC and approval checks never ran. The harm continued afterwards through leaked borrower data and repeat attacks.
The money was only the first loss. Borrower data from these breaches can now fuel fraud calls, and some lenders were hit a second time because the first hole was not fully closed. This article walks through the whole chain, from the first weakness to the harm that lands weeks later, and shows where each link can be broken.
One caution before we start. The ₹300 crore figure comes from unnamed sources, and no regulator has confirmed it. We label each fact below as reported or described in an official advisory, so you can see how firm it is.
Key facts: confirmed vs reported
Nothing in this table is an official total. The status column shows where each fact comes from.
| Fact | Detail | Status |
|---|---|---|
| Estimated loss | About ₹300 crore | Reported estimate, not an official total |
| Period | Roughly two months, intensifying in the last six weeks | Reported |
| Complaints and FIRs | Close to 150 | Reported |
| Most affected | NBFCs with annual turnover of about ₹1,000 crore or less | Reported |
| Attack path | Stolen payment API credentials used from the lender's own whitelisted servers | Described in CERT-In advisory, 9 Oct 2026 |
| Fastest case | Hyderabad NBFC: ₹3.3 crore, 183 transfers, 52 accounts, 87 minutes | Reported by police |
| Repeat case | Delhi NBFC: ₹7.5 crore, then ₹7.25 crore days later | Reported |
| AI involvement | Some completed probes linked to AI hacking toolkits | Reported; role in each case not established |
| Affected lenders named | None officially | Not disclosed |
The seven stages of the NBFC attacks
A cyberattack on a lender is not one event. It is a chain that runs from the first exposed weakness to harm that lands long after the money has gone. Stages 1 to 5 follow the pattern CERT-In described. Stages 6 and 7 begin after the money leaves, which is where most incident response stops.
Stage 1: Exposure
Every attack starts with something reachable that should not be. It might be a payment credential sitting in a configuration file, a deprecated API that is still online, or an unpatched internet-facing application. CERT-In asks financial entities to keep a complete API inventory and to decommission unused APIs, because this is where the chain begins.
Stage 2: Compromise
Attackers break into internet-facing applications and exposed APIs, then plant web shells and backdoors so they can come back. Reporting indicates they studied their targets for 10 to 12 days first. They mapped payment flows, whitelisted IDs and weaknesses in application code before they acted.
Stage 3: Trusted-channel abuse
This is the stage that explains how KYC was bypassed. With stolen credentials, attackers call disbursal and fund-transfer APIs directly from the lender's own whitelisted infrastructure. Loan creation, KYC, sanction, beneficiary validation and per-transaction limits all live inside the lending application, so none of them is ever invoked.
The request is authenticated. Nothing checks whether it is legitimate.
Stage 4: Execution
Money moves in many small transfers sized to look routine. In the Hyderabad case, police reported 183 transfers of ₹50,000 to ₹2 lakh to 52 accounts in 87 minutes. That pace is automation, not a person at a keyboard, and it should have tripped a velocity control.
Two other cases show the same shape. A Delhi Police Special Cell case, reported by a single source, involved malware on an NBFC server and ₹3.93 crore in 149 transfers on 5 October, with seven arrests. The earlier Moneyview case involved about ₹49 crore in about 1,782 transfers. The Delhi case has not been independently verified, and its route (malware on a server) differs from the stolen-API-key pattern in CERT-In's advisory.
Stage 5: Loss
Funds pass quickly through mule accounts. In Hyderabad, police could block only about ₹40 lakh of the ₹3.3 crore. Across the sector, the loss is estimated at about ₹300 crore.
Stage 6: Spillover to borrowers
Officials quoted in the reporting warn that data taken in these breaches could fuel Jamtara-style fraud calls. A caller who knows a borrower's name, lender and loan amount sounds highly convincing, and the borrower usually has no idea a breach occurred. This harm is rarely counted in an incident's total.
Stage 7: Recurrence
A Delhi NBFC reportedly lost ₹7.5 crore and then ₹7.25 crore a few days later. Reporting suggests that firms that had not fully fixed their weaknesses were hit again. The second loss is not a detection failure, because the breach was already known. It is a failure to close every path the attacker had.
Why machine identities are the weakest link
Payment API keys are the most powerful identities inside a digital lender, and the least protected. Human logins get multi-factor authentication, device checks and access reviews. A payout credential can move crores with none of those, because no person ever logs in.
These credentials spread across configuration files, CI pipelines, scripts, repository history and vendor platforms. Loan origination systems, CRMs and collections tools often hold copies. Each copy is a separate route to the money, and each one must be found, rotated and revoked after a breach.
This is why one stolen key can be worth more than a stolen employee password. It needs no phone, no OTP and no office hours.
Were the NBFC attacks carried out by AI?
They were AI assisted, not AI run. Investigators linked some completed probes to AI hacking toolkits, and CERT-In says such tools cut the time from break-in to damage from days to hours. The techniques themselves (credential theft, API abuse, mule accounts) are long established.
What AI changes is cost and reach. Reconnaissance that once took a skilled operator weeks can be partly automated, so weaknesses that were once too costly to find are now worth exploiting. The precise role of AI in each case has not been established, and claims that autonomous agents ran the attacks are not supported by the reporting.
The practical lesson is simple. Defences built for slow, manual attackers, such as quarterly reviews and weekly log checks, are too slow for this pace.
What CERT-In and RBI have said
CERT-In and CSIRT-Fin issued a critical advisory (CIAD-2026-0046) on 9 October 2026. It covers NBFCs, digital lenders, fintechs, payment service providers and wallet operators. It recommends multi-factor authentication, a full API inventory, decommissioning unused APIs, per-session authentication, credential rotation and maker-checker controls.
A second CERT-In advisory on 15 October flagged AI-based tools used for reconnaissance and exploitation. On 19 October, the Reserve Bank of India asked regulated entities to treat their environment as compromised wherever credentials are exposed or unauthorised access is confirmed. That instruction speaks directly to stage 7, because it tells lenders not to assume the first fix was enough.
None of these advisories names an affected lender, a threat actor, an incident count or an amount lost. The loss figure in this article comes from press reports, not from them.
Why smaller NBFCs were hit hardest
Smaller NBFCs digitised lending quickly but rarely funded security at the same pace. Most affected lenders have turnover of about ₹1,000 crore or less, small security teams and long audit cycles.
They are also deeply interconnected through APIs, co-lending arrangements and UPI collections. A weakness at a small lender reaches its partner banks, and its exposure includes every vendor that holds its keys or borrower data. An attacker does not need to beat the biggest bank. They only need the weakest connected lender.
For the full control set, see our DPDP and RBI digital lending implementation guide for NBFCs, and the pages on digital lending and loan apps and lending service providers.
What the DPDP Act adds
The stolen money is a fraud and security problem. The borrower data that follows is a data protection problem, and that is where the DPDP Act comes in. Three points matter for lenders.
- The breach duties are not in force yet. The Act's section 8 duties and Rule 7 of the DPDP Rules start on 13 May 2027. Today, lenders report to CERT-In and the RBI under existing rules, which MediaNama reports include a 6-hour window.
- The future duty is two-step. Under section 8(6) and Rule 7, a Data Fiduciary must tell each affected person without delay. It must also tell the Data Protection Board without delay, then send a detailed report within 72 hours, unless the Board allows longer in writing.
- Weak safeguards carry the highest penalty. The maximum is ₹250 crore for failing to keep reasonable security safeguards under section 8(5), and ₹200 crore for failing to give breach notice under section 8(6). These are ceilings, and a penalty follows only after a Board inquiry finds a significant breach and the firm has been heard.
One open question deserves honesty. The reported incidents are about stolen funds moving through payment APIs. It is not yet clear how many of them exposed personal data. Where borrower data was taken, as officials warn, the same stolen-key weakness would also be a failure of the security safeguards the Act expects. Rule 6 lists those safeguards: encryption or masking, access control, logging and monitoring, backups, and processor contract terms. Our DPDP Act vs ISO 27001 comparison shows how far a certification goes towards these safeguards.
For a lender, the preparation is the same under every regime: know where your credentials and borrower data sit, and be able to say within hours, not weeks, what was touched. Our DPDP Rules guide and breach notification glossary entry cover the duties in more detail.
Lenders that use cloud or payment vendors should also check where data may be stored and sent. Our cross-border data transfer guide and sector data localisation rules (RBI, SEBI, IRDAI) explain how the DPDP Act sits beside those rules.
Where the chain can be broken
Every stage has a control that can stop the chain, and every control needs a named owner. Gaps appear where ownership passes from one team to the next.
| Stage | Control | Typical owner |
|---|---|---|
| 1. Exposure | API inventory; remove credentials from code and config; decommission unused APIs | Technology / security |
| 2. Compromise | MFA on admin and payment platforms; patch internet-facing apps; monitor for web shells | Security |
| 3. Trusted-channel abuse | Per-session authentication; bind payouts to sanctioned loan IDs | Technology + operations |
| 4. Execution | Velocity limits and circuit breakers on payout APIs; cooling-off for new beneficiaries; independent approval channel | Fraud + operations |
| 5. Loss | Fast reporting to CERT-In and the partner bank; fund tracing | Fraud + incident response |
| 6. Spillover | Notify affected borrowers; warn about impersonation calls | Customer operations + compliance |
| 7. Recurrence | Revoke every copy of every exposed credential, remove backdoors and verify all paths closed before declaring the incident over | Board-designated owner |
If you can only do one thing first, bind every payout to a sanctioned loan ID (stage 3). It turns a stolen key from a blank cheque into a key that only opens doors that already exist.
To test where your own controls stand, try the free DPDP readiness assessment, or read about our gap assessment service and privacy audit services.
Why information sharing matters
No affected lender has been publicly named, and none has shared what happened. The legal and reputational reasons are clear, but the effect is that peers learn from estimates rather than from evidence.
Confidential, structured sharing of indicators and attack patterns is how a campaign against one lender becomes a warning to the next. Smaller NBFCs need it most, because they have the smallest security teams and the least chance of spotting the pattern alone.
What borrowers should do
If you have a loan with an NBFC, assume a caller may know your details.
- Never share OTPs, PINs or passwords, and never move money on a call.
- Hang up and call the lender back on the number on its official website or your loan documents.
- Report suspected fraud to the national cybercrime helpline 1930 or at cybercrime.gov.in.
Frequently asked questions
How much money did NBFCs lose in the 2026 cyberattacks?
Which NBFCs were hacked?
Were the NBFC attacks done by AI?
What is CERT-In advisory CIAD-2026-0046?
How did attackers bypass KYC and loan approvals?
Why was the same NBFC hacked twice?
Is my data at risk if my NBFC was attacked?
How can NBFCs stop unauthorised payouts?
Sources
- The Economic Times, 29 Oct 2026, via LinkedIn News and NewsBytes
- MediaNama: AI cyberattacks on NBFCs, Rs. 300 Cr lost, disclosure gaps, 30 Oct 2026
- CERT-In and CSIRT-Fin advisory CIAD-2026-0046: Ampcus Cyber summary, VARINDIA
- CERT-In 15 Oct advisory and RBI 19 Oct advisory: Kotak Neo
- Hyderabad case: Siasat
- Delhi repeat case: The420.in
- Deccan Herald: Moneyview case, Rs 49 crore
- Elets BFSI: Delhi Police Special Cell case (single source, not independently verified)
- RMA India: Hyderabad NBFC case
- Digital Personal Data Protection Act 2023 (sections 8(5), 8(6), 33) and DPDP Rules 2025 (Rules 6 and 7), as published in the Gazette of India
Last checked: 30 October 2026. Loss figures and case details are media-reported and have not been confirmed by the RBI or CERT-In. The DPDP section is general information, not legal advice. Confirm your obligations with counsel.