Incident sheet · /security

Filed
Read
4 min · 751 words
Severity
Routine

AI-assisted hackers diverted Rs 65 crore in approved loans from 12 NBFCs

AI-assisted hackers diverted over Rs 65 crore in approved loans from 12 Delhi-NCR NBFCs by hijacking disbursal APIs through whitelisted cloud IPs, Delhi Police say.

AI-assisted hackers breach NBFCs in Delhi-NCR; divert Rs 65 cr in approved loans
FeedAI-assisted hackers breach NBFCs in Delhi-NCR; divert Rs 65 cr in approved loans — AI-generated

Event log

  1. Over Rs 65 crore diverted from 12 NBFCs in Delhi-NCR in three months, part of 150+ FIRs across India.
  2. Attackers redirected approved, undisbursed loans to mule accounts via disbursal APIs using whitelisted cloud IP addresses.
  3. Rs 18 crore seized; 20 people arrested, all tied to mule accounts.
  4. RBI issued two confidential advisories, including one on API hijacking on November 16.
  5. CERT-In advisories on October 9 and 15 omit affected entities, threat actors and loss figures.

Hackers, apparently assisted by AI, diverted more than Rs 65 crore in approved-but-undisbursed loans from 12 non-banking financial companies across Delhi-NCR in three months. Senior Delhi Police officers told The Indian Express that these cases form part of over 150 first information reports filed by NBFCs across India.

The attackers skipped the fashionable "digital arrest" scam entirely. Instead they hunted for loans that lenders had already approved but not yet paid out, then redirected the disbursements into mule accounts.

"We have detected traces of AI being used in these frauds. These incidents are largely attributable to weaknesses in the public-facing cyber systems of NBFCs," said Gaurav Tyagi, Deputy Commissioner of Police at the Intelligence Fusion and Strategic Operations unit.

How did the attackers get in?

Police describe a plain malware infection at the front end. Many of the smaller firms targeted carry flaws in their software code, and the malware sat inside their systems for seven to eight days before anyone noticed.

The disbursement trick is equally unsophisticated in principle. A typical NBFC hires a fintech to host its website on a cloud data centre — Amazon Web Services, for instance. The firewall whitelists the data centre's outbound IP addresses, and a disbursal request is approved only if it arrives from one of them.

Anyone can open an account on those cloud platforms. Fraudsters therefore sent commands from addresses the NBFC's firewall already trusted, instructing the API that links the lender to its bank to release a loan — with a mule account substituted for the borrower's. The bank approved it instantly.

Delhi Police have arrested 20 people across the 12 NCR cases, all linked to creating or procuring mule accounts. Investigators seized Rs 18 crore of the Rs 65 crore before it left those accounts as cash or crypto.

Which companies lost money?

The Indian Express named only two victims. Hero FinCorp reportedly lost Rs 12 crore on August 17. Unifinz Pvt Ltd claimed a loss of Rs 12.49 crore across 550 transactions over two days in a complaint dated October 21. The remaining entities stay unnamed; their losses range from Rs 66 lakh in 66 transactions to Rs 10.59 crore in 1,131 transactions, occurring between September 18 and November 3.

What has the regulator done?

Plenty — mostly in confidence. In November, the Reserve Bank of India summoned banks over the rise of AI-assisted cyber-attacks on the financial sector and, according to reports, delivered a "dressing down" before issuing a confidential advisory on securing systems. A second confidential advisory followed on November 16, addressing API hijacking.

CERT-In has been busier on paper:

  • October 9: CERT-In and CSIRT-Fin issued a "Critical" advisory to NBFCs, digital lenders, fintechs, payment service providers and wallet operators, describing how attackers steal payment gateway and partner-bank API credentials to reach disbursal APIs directly — bypassing KYC checks and per-transaction limits.
  • October 15: A second advisory warned that financially motivated attackers were running campaigns against Indian entities using publicly available, AI-native offensive security frameworks.
  • October 27: CERT-In ran a full-day drill with more than 300 officials from over 55 financial institutions, mostly regional rural and cooperative banks.

None of the advisories names an affected entity, identifies threat actors or tools, or discloses incident counts and stolen sums from CERT-In's own records. Sanjay Bhatia, Additional Commissioner of Police at Delhi Police's IFSO, asked financial entities to follow the CERT-In advisories on AI-driven attacks.

Why did the public learn this from the press?

The pattern is consistent. In October, Economic Times reported that AI-powered attacks had siphoned roughly Rs 300 crore from smaller NBFCs nationwide over two months, producing more than 150 FIRs — a figure regulators never published.

Central Bank of India's Rs 13.97 crore loss surfaced through a Hindustan Times report on its police complaint before the bank told the National Stock Exchange about the network vulnerability, without mentioning the money allegedly lost. SBI informed stock exchanges about a data exposure two days after a dark web claim appeared. CoinDCX's chief executive called breach reports a false alarm, then admitted a Rs 152 crore loss three days later.

What does this change for CIOs at lending firms? If your disbursal API trusts any traffic born on a shared cloud provider's IP range, that trust is now a documented liability — and the regulator will not be the one telling your customers.

via medianama.com (Source)

Filed under

  • ai-assisted-cyberattacks
  • nbfc-security
  • loan-fraud
  • api-security
  • cloud-security
Share this article:

More from Tom Whitfield

Tom Whitfield

Show full bio

Market editor covering people, launches and day-to-day practice in enterprise information technology and the leaders who run at CIO Monitor.

185 articles

The newsroom digest

One email a day with the stories that matter. No noise.

By subscribing you agree to the privacy policy.

Cross-references

‹ Previous articleNext article ›