It is not just a phishing email with better grammar. It is a machine that scans your bank’s APIs, writes its own attack scripts, tests a ₹1 transfer to see if the door is open, and waits for a weekend when the security team is thin. India’s banks are not facing a wave of smarter criminals. They are facing a new class of adversary that never sleeps, never specialises, and never attacks the same way twice. This is the concept behind AI-powered cyberattacks — not the patch notes, not the press releases, just the idea that offence has quietly become automated, and defence has not caught up.
Close your eyes and picture a bank on a Sunday morning.
The building is quiet. The security operations centre has a skeleton crew. Somewhere in a cloud region, a process is running that nobody is watching closely. It tries a ₹1 transfer to a newly added beneficiary. It works. Then it tries again, and again, each time slightly different, each time learning from the response. By Monday morning, the money is gone, the logs look almost normal, and the team is trying to reconstruct what happened from fragments.
That is not a hypothetical. That is the playbook CERT-In flagged in November 2026, and it is the reason the phrase “AI-powered cyberattack” has moved from conference panels to actual advisories. Once you understand that idea — offence at machine speed, defence still built for human speed — you understand why Indian banks are suddenly rethinking everything from API security to weekend staffing.
The Advisory That Made Everyone Pay Attention
On November 4, 2026, India’s cybersecurity watchdog, the Indian Computer Emergency Response Team (CERT-In), issued an advisory identifying a threat actor called “Crafty Falcon” — an AI-assisted financial threat actor using publicly available AI tools to execute attacks autonomously and at machine speed[reference:0]. The campaign was already an escalation of an earlier warning from October 9, 2026 about attackers exploiting payment and disbursement APIs to move money without authorisation[reference:1].
The detail that made the advisory different from the usual boilerplate was this: CERT-In said the threat was active and ongoing, and that successful attacks were still being reported — including from NBFCs and fintech entities that had already received the earlier warning and had undertaken remediation[reference:2]. Some organisations were deliberately targeted again, including those that had not fully fixed the vulnerabilities identified earlier[reference:3].
That is the part banks found uncomfortable. Closing the initial entry point was not enough. The attackers adapted, came back, and changed their approach when the first one was blocked.
The threat is not that attackers have better tools. It is that they can now use those tools without needing to understand what the tools are doing. That is a different kind of adversary.
The Tools: AI That Writes the Attack While It Runs
CERT-In did not just name the threat actor; it named the tools. The advisory identified several publicly available AI-driven offensive security platforms being used in the campaign[reference:4]:
ARTEX, a family of large language model-based tools that can generate and refine attack scripts; PentAGI, an autonomous multi-agent AI penetration-testing system; Cairn, an attack search engine; and CyberStrikeAI, a framework for automating offensive workflows[reference:5]. According to the advisory, these tools can execute entire attack chains autonomously, generate customised scripts in rapid succession, and adapt their payloads based on how a target system responds[reference:6].
That last clause is the important one. Adaptation is the difference between a script and an intelligent attack. A traditional exploit does the same thing every time. These tools read the response, adjust, and try again — at a scale no human team could match.
CERT-In also observed signs of machine-generated coding: attackers producing large numbers of functionally similar scripts and modifying them in quick succession[reference:7]. This is not a lone hacker in a basement. It is an industrial process, running on infrastructure that rotates IP addresses within hours to days, making conventional detection unreliable[reference:8].
How the Money Actually Moves
The attack surface is not the bank’s core ledger. It is the seams — the APIs that connect banks to payment aggregators, NBFCs, fintech apps, and cloud services.
CERT-In identified several methods. Attackers exploited internet-facing testing and staging environments to reach production systems through shared credentials, tokens, and network connections[reference:9]. They abused cloud services such as AWS Systems Manager and Lambda to execute commands and deploy malicious files while blending in with legitimate administration[reference:10]. They reverse-engineered application code and used modified app packages to make unauthorised API calls, defeating controls that assume a request must come from the genuine app[reference:11].
And then there is the ₹1 transaction. Attackers used micro-value transfers, sometimes as low as one rupee, to test whether an unauthorised payment route was working before moving larger sums[reference:12]. Merchant settlement APIs of payment aggregators were also exploited to divert funds[reference:13]. The pattern is consistent: find a seam, test it quietly, then scale.
The advisory also flagged newly disclosed vulnerabilities in PHP and Node.js applications, and warned that cryptocurrency exchanges and virtual digital asset providers were being targeted for wallet keys and funds[reference:14]. The perimeter is not a wall. It is a mesh of third-party services, and every node is a potential entry point.
The Numbers Behind the Fear
This is not a niche concern. India’s BFSI sector suffered 2.9 million cyberattacks in 2025, more than double the 1.4 million recorded in 2021, according to the government’s Digital Threat Report 2025–26[reference:15]. The report describes an “AI asymmetry”: offensive capabilities are scaling faster than the regulatory, defensive, and operational frameworks designed to contain them[reference:16].
A BioCatch survey of banking leaders found that 84% of Indian respondents reported rising fraud losses over the past year, and 93% believe AI has increased the sophistication of fraud and scam schemes[reference:17][reference:18]. Nearly 84% identified AI agents as the industry’s greatest exploitable vulnerability in the coming year[reference:19]. And 43% of CISOs surveyed said attackers had already outpaced their defences, while only 19% had raised security budgets by more than 10%[reference:20].
The uncomfortable arithmetic is simple: attackers are automating faster than defenders are hiring. A bank can add a few analysts. An AI attacker can add a thousand parallel attempts.
The RBI’s Response: From Principles to Proof
The Reserve Bank of India is not sitting still. In its FY27 agenda, the central bank has made AI-related cyber risk a first-class supervisory concern, including plans for an AI innovation sandbox, new norms for AI use in finance, and a cyber-mapping exercise for NBFCs[reference:21][reference:22].
More concretely, the RBI has begun asking banks to document how they meet cybersecurity rules and describe their plans for AI-related risks, with a specific focus on API weaknesses that connect banks to fintechs and non-bank lenders[reference:23]. A consultant quoted in the report said some lenders had yet to fully implement the RBI’s existing cybersecurity directions — meaning the gap is not just technology, but execution[reference:24].
CERT-In’s recommendations read like a checklist for the next twelve months: round-the-clock monitoring with extra staffing on weekends and holidays, isolation of testing and staging environments, server-side transaction validation, API request signing, app attestation, faster patching of internet-facing applications, identity-based access controls instead of IP allow-lists, and zero-trust segmentation[reference:25]. It also advised restricting unauthorised AI-driven security tools on corporate systems and monitoring LLM API usage — a quiet acknowledgment that the same AI capability that helps attackers can be misused inside the enterprise[reference:26].
Why This Should Stay With You
The next time you see a bank’s cybersecurity page and it talks about firewalls, encryption, and 24/7 monitoring, remember what is actually happening underneath. The threat is not a single breach. It is a continuous, automated search for the weakest seam in a mesh of APIs, cloud services, and third-party integrations.
India’s digital payments are projected to reach $3.1 trillion by 2028, representing 35% of total banking revenue — a growth story regulators want, but one that expands the attack surface faster than compliance frameworks adapt[reference:27]. The question is not whether AI will be used in attacks. It already is. The question is whether banks can move from documenting controls to operating them at the same speed.
That is the concept of AI-powered cyberattacks on Indian banks. Not a story about criminals getting smarter, but a story about offence becoming automated — and defence still being measured in meetings, approvals, and quarterly reviews. The gap between those two speeds is where the next incident will happen.
Fact-check note: CERT-In issued an advisory on November 4, 2026, identifying the “Crafty Falcon” campaign as an AI-assisted threat actor exploiting payment and disbursement APIs. The advisory named tools including ARTEX, PentAGI, Cairn, and CyberStrikeAI, and described techniques including micro-value transactions (as low as ₹1), testing-environment exploitation, cloud-service abuse, and modified application packages. India’s BFSI sector recorded 2.9 million cyberattacks in 2025, up from 1.4 million in 2021, per the Digital Threat Report 2025–26. The BioCatch survey figures (84% reporting rising fraud losses, 93% saying AI increased fraud sophistication, 84% flagging AI agents as the greatest exploitable vulnerability) are based on a survey of 100 Indian banking and financial crime prevention leaders. RBI’s FY27 agenda includes AI risk detection, an AI innovation sandbox, and demands for banks to document AI-related cyber preparedness. Sources: Business Standard, MoneyLife, Inc42, MediaNama, The Hindu BusinessLine, WION, IndiaBank.net, CERT-In advisories.
Image credits: All images are from Pexels and are free to use under the Pexels License. Representative images have been used for illustrative purposes; no official CERT-In, RBI, or bank assets are included. Fig. 1: Mikhail Nilov; Fig. 2: Tara Winstead; Fig. 3: Brett Sayles; Fig. 4: Pixabay; Fig. 5: Karolina Grabowska.
Login with