Cybersecurity Agency I4C Warns Banks & Fintechs Against AI-Led Cyberattacks

India’s I4C warned banks, non-banking financial companies and fintech startups about attackers using AI agents for financial fraud. The National Cybercrime Threat Analysis Unit said attackers were exploiting unpatched software and misconfigured internet-facing servers to enter financial infrastructure. Customer-facing complaint interfaces may also provide entry points.

After gaining access, attackers can target automated fund-disbursement systems and payment or transfer APIs. They may route money through mule accounts across banks, withdraw it quickly through IMPS and NEFT, then delete records. The agency reported multiple individual cases involving unauthorised transactions running into crores of rupees.

The warning matters because AI can make attacks faster and more automated. I4C advised institutions to patch systems, test APIs, set transaction limits, verify beneficiaries and monitor unusual activity. It also called for six-hour reporting to CERT-In, stronger coordination and authorised red-team exercises against machine-speed attacks.

An agentic AI system is software that can work toward a goal with limited human intervention. Unlike a text generator, it can inspect information, choose the next step, call connected tools or APIs and repeat the process. Its ability to act depends on the permissions and systems connected to it. The article does not provide a technical definition, but describes malicious agentic AI systems operating inside compromised financial systems.

For example, an attacker could use an agent to navigate an exposed system, identify an automated disbursement process and send unauthorised payment instructions. It could then help route funds through mule accounts and support rapid withdrawals through IMPS or NEFT. These actions are possible only when the system has access and sufficient permissions.

This makes agentic AI a serious security concern for financial institutions. I4C urged firms to apply CERT-In advisories, test payment APIs, enforce KYC and approval checks, monitor transaction patterns and conduct authorised red-team exercises against machine-speed attacks.

Unpatched software may contain known weaknesses that attackers can exploit because fixes have not been installed. Misconfigured internet-facing servers can also expose financial infrastructure to unwanted access. Customer-facing complaint interfaces are another possible entry point, especially when they connect to internal systems or accept data that is handled insecurely. I4C highlighted all three risks.

Once inside, attackers may move toward automated fund-disbursement systems and APIs used for payments and transfers. The key danger is the connection between an initial access point and systems that can trigger financial actions. If controls are weak, a compromise can become an unauthorised transaction rather than remain a limited technical intrusion.

The agency advised financial institutions to implement critical CERT-In advisories urgently, test payment APIs and strengthen customer-facing infrastructure. It also recommended transaction limits, beneficiary verification, KYC and approval checks, monitoring of unusual transaction patterns and authorised red-team exercises to test response speed.

I4C reported multiple individual cases in which malicious agentic AI systems exploited compromised financial systems for unauthorised transactions running into crores of rupees. The article does not state the exact number of cases, the exact total or the amount in each case. Therefore, the scale is significant, but it cannot be reduced to one precise figure from the advisory.

In Indian numbering, one crore rupee equals Rs 10 million, or Rs 1,00,000. “Crores of rupees” means the reported amounts reached several units of that scale. In practical terms, this describes transactions involving many tens of millions of rupees, rather than a small loss. The wording indicates magnitude, not a confirmed total.

The financial scale helps explain the warning’s urgency. Attackers allegedly targeted automated disbursement systems and payment APIs, then moved funds through mule accounts and withdrew them quickly. I4C responded by recommending transaction limits, beneficiary checks, monitoring and stronger approval controls before transfers.

Mule accounts are accounts used to receive or pass on criminal proceeds. According to I4C, attackers route stolen funds through mule accounts across different banks, then quickly withdraw the proceeds through IMPS and NEFT. This creates a fast-moving chain between the compromised system, receiving accounts and cash-out activity.

The mechanism combines speed and separation. Money does not remain in the original destination account, where it may be easier to identify or block. It moves through multiple accounts and payment channels, while rapid withdrawals reduce the time available for intervention. The perpetrators then delete records and hide their tracks, according to the NCTAU.

The consequence is a more difficult investigation and a higher risk of irreversible loss. I4C recommended transaction limits, beneficiary verification, monitoring for sudden changes in speed, volume and patterns, and completed KYC and approval checks. Better coordination between security, transaction-monitoring and IT teams can also improve detection and response.

Financial institutions can add friction at the points where an automated attack tries to move money. I4C recommended transaction limits, beneficiary verification and monitoring for sudden changes in transaction speed, volume and patterns. KYC and approval checks should be completed before funds are transferred. These controls can help stop suspicious instructions before they become losses.

Technical defenses matter too. Institutions should implement CERT-In’s critical advisories on payment and disbursement APIs and AI agents, test payment APIs and strengthen customer-facing infrastructure. They should also coordinate security, transaction-monitoring and IT teams, since an attack may begin as a technical breach but quickly become a transaction problem.

Preparation is part of the defense. I4C advised authorised red-team exercises to test how quickly security and payment controls respond to machine-speed attacks. Firms must report cyber incidents to CERT-In within six hours, and banks or NBFCs must also report them to RBI. Together, these steps improve prevention, detection and response.

An API, or application programming interface, is a defined way for one software system to communicate with another. Banks and fintechs use APIs to connect applications, process payments, transfer funds and operate automated disbursement systems. They make digital financial services faster because systems can exchange instructions without manual handling for every transaction.

The same connection creates risk when an API is exposed, misconfigured or reached after another system is compromised. An attacker may target payment or transfer APIs and send unauthorised instructions, especially if transaction limits, beneficiary checks or approval gates are weak. In the article’s cases, attackers potentially targeted automated fund-disbursement systems and APIs after entering financial infrastructure.

I4C advised institutions to implement critical CERT-In guidance on payment and disbursement APIs, test those APIs and strengthen customer-facing infrastructure. It also recommended monitoring transaction changes, completing KYC and approvals, and using red-team exercises to test defenses against machine-speed attacks.