India's I4C and CERT-In warned AI-agent attacks on banks and NBFCs led to multi-crore unauthorised transactions
India's Indian Cyber Crime Coordination Centre (I4C) and CERT-In issued a joint advisory saying cybercriminals are using AI agents to compromise financial institutions at machine speed. Attackers exploit unpatched software, misconfigured servers and payment APIs, move funds to mule accounts, and then delete records to hide their tracks; multi-crore unauthorised transactions were found in individual cases. The agencies did not name the affected banks or institutions. They urged banks, NBFCs and fintech firms to apply critical-severity CERT-In measures and report incidents to CERT-In and the RBI within six hours.
What the AI did
Cybercriminals used AI agents, software that can carry out a series of steps on its own, to attack banks and other financial institutions at machine speed. The attackers exploited unpatched software, misconfigured servers and payment APIs (the connections that let systems send payments to each other), moved money to mule accounts and deleted records to hide their tracks. It has not been reported which AI tool was used.
First reported November 4, 2026 · Added to the register November 10, 2026 · 4 sources
- Alleged loss by two Mumbai lenders
- approximately $2 million
- Model
- ARTEX, PentAGI, Cairn, CyberStrikeAI
- Affected
- Banks, NBFCs and fintech firms, NBFCs and fintech companies
- Country
- India
- When it happened
- Not stated in the sources
- First reported
- November 4, 2026
What this means for you
Could this affect you?
If you are a bank, NBFC or fintech firm in India. Indian banks, NBFCs and fintech firms were warned that attackers using AI agents target automated fund disbursement systems and payment APIs, moving money to mule accounts. Multi-crore unauthorised transactions have been found in individual cases.
What to check
- Inventory all payment-related APIs and disable obsolete, unused and testing interfaces.
- Strengthen monitoring of automated fund disbursement systems.
- Patch all internet-facing systems.
- Prepare your security, transaction monitoring and IT teams to report incidents to CERT-In within six hours.
Areas of your AI programme this touches
Timeline
- November 4, 2026First reported
- November 5, 2026In an advisory issued on November 4, 2026, CERT-In named the group behind the attacks 'Crafty Falcon' and named four AI tools it used: ARTEX, PentAGI, Cairn and CyberStrikeAI, at least two of which security researchers have linked to China. CERT-In did not say who was behind the group or which incidents were linked to it.[4]
- November 5, 2026CERT-In said the campaign has involved testing payment routes through transfers as small as ₹1 before attempting larger unauthorised payments. It said successful attacks have continued, including against NBFCs and fintech companies that had already attempted remediation, some of which were targeted again after incomplete security fixes.[2]
- November 5, 2026Citing a police complaint, Reuters previously reported that two Mumbai-based non-banking lenders lost approximately $2 million through more than 700 unauthorised instant transfers on October 8 and 9, handled through ICICI Bank. The advisory follows a series of cyber incidents affecting Indian financial institutions in recent weeks.[3]
Every fact and its source (15)
- Advisory issued2026-11-04“according to an advisory issued on Wednesday (November 4, 2026)”[4]
- Losses foundMulti-crore unauthorised transactions in individual cases“Multi-crore unauthorised transactions through fund disbursement systems and APIs are being found in individual cases”[1]
- Victims unnamedNo specific banks identified“The agency has not identified specific banks or financial institutions affected by these incidents.”[1]
- AI methodAgentic AI framework enabling machine-speed attacks“Use of Agentic AI Framework makes the attack possible at machine speed”[1]
- Cover-upAttackers delete records to hinder investigation“The perpetrators delete records and hide their tracks so that investigation becomes difficult”[1]
- Fund destinationMule accounts across multiple banks“The money is subsequently transferred to 'mule' accounts”[1]
- ResponseIncidents to be reported to CERT-In within six hours“report such incidents to CERT-In within six hours of detecting them”[1]
- Timing tacticAttacks timed around holidays and weekends“Attacks were timed around holidays, weekends and other periods where monitoring was likely to have been low to delay detection.”[4]
- Threat group named by CERT-InCrafty Falcon“calls “Crafty Falcon,” has breached non-bank lenders, fintech firms”[4]
- MethodMoves money using victims' own payment and disbursement systems“moves money using victims’ own payment and disbursement systems”[4]
- Undisclosed detailsCERT-In did not name affected firms or amount stolen“CERT-In did not name the affected companies or say how much money was stolen.”[4]
- Funds stolen fromMerchant settlement APIs and crypto wallets“Attackers have also targeted merchant settlement APIs and cryptocurrency wallets, stealing wallet keys and siphoning funds.”[2]
- Re-targetingFirms hit again after incomplete fixes, on weekends and holidays“Some were targeted again after incomplete security fixes, with attackers timing operations around weekends and holidays.”[2]
- Technique₹1 test transfers before larger unauthorised payments“The campaign has involved testing payment routes through micro-value transactions, as small as ₹1 transfers, before attempting larger unauthorised payments.”[2]
- Bank that handled the transfersICICI Bank“The deals were handled through ICICI Bank”[3]
Sources
- AI frauds on banks & NBFCs: Central cyber security agency warns banks, NBFCs of AI attacks; I4C reveals how hackers are executing multi-crore fraudseconomictimes.indiatimes.com · November 4, 2026
- AI hackers use ₹1 test transactions to move bigger sums from financial firms: CERT In warnsbusinesstoday.in · November 5, 2026
- CERT-In flags China-Linked AI cyberattacks on Indian banks: Hackers exploit holidays, low-monitoring windowslivemint.com · November 5, 2026
- Hackers used China-linked AI tools to steal from Indian financial firmsthehindu.com · November 5, 2026
How this record is classified. Severity S (4/5): real harm to one organisation, data exposed, or a containment failure. OECD level: incident, an event in which an AI system led to actual harm. Evidence: Confirmed, meaning a company, official or court statement.
The facts in At a glance and in Every fact and its source are quoted from the sources listed. The summary, What the AI did and What this means for you are written from those sources and checked against them automatically before publication. Records reflect what has been disclosed, not everything that has happened. Incidents are found through news feeds and the GDELT Project.
