Hackers used China-linked AI tools to steal from Indian financial firms

Indian financial firms faced a coordinated wave of intrusions that targeted non-bank lenders, fintech companies and cryptocurrency exchanges. The attacks mattered because they went beyond stealing data: they also reached systems used to approve and move money. CERT-In called the threat active and ongoing, while authorities increased warnings and the Reserve Bank of India tightened cybersecurity checks.

The attackers reportedly used weaknesses in application programming interfaces, or APIs, which connect banks with lenders and fintech firms. They sent loan-disbursal instructions through those links and used victims’ own payment and disbursement systems. Attacks were timed for holidays and weekends, when monitoring could be weaker. Firms were also attacked again after incomplete fixes.

The full list of affected companies remains unknown. CERT-In did not identify linked incidents, disclose total losses, or say who was behind the breaches. Other reported incidents included a dark-web leak of 3,897 State Bank of India customer records and unauthorised debits reported by Spandana Sphoorty Financial.

Crafty Falcon is a hacking group identified by India’s Computer Emergency Response Team, or CERT-In. The agency used that name for attackers who breached non-bank lenders, fintech firms and cryptocurrency exchanges. Identifying the group matters because it gives financial institutions a way to connect separate incidents and focus defensive measures, even though the group’s identity and location remain unknown.

CERT-In said Crafty Falcon surveys targets with machine intelligence and automates break-ins “at machine speed.” The group then uses victims’ own payment and disbursement systems to move money. It also chose holidays, weekends and other low-monitoring periods, and returned to firms where earlier security fixes were incomplete.

CERT-In called the threat active and ongoing. However, it did not say which reported incidents were caused by Crafty Falcon, who operated the organisation, or whether the attacks were state-sponsored. The named tools were freely available online, so the group’s exact affiliation remains unconfirmed.

CERT-In named four AI tools used by the attackers: ARTEX, PentAGI, Cairn and CyberStrikeAI. Their importance lies in how publicly available software can help automate reconnaissance and intrusions against financial institutions. The agency did not say that the tools themselves prove who operated the attacks, and it did not identify the people or organisation behind Crafty Falcon.

Security researchers have linked at least two of the tools to China. CrowdStrike said China-made ARTEX was used against South Korean banks from late September. It also reported that a China-linked group used ARTEX in an espionage campaign targeting more than 200 organisations across at least 14 countries. ARTEX was made private by its developer last month after the South Korean attacks.

These links provide evidence about some tools, not definitive proof that the Indian attacks were state-sponsored. CERT-In did not make that claim. The other tools’ specific origins or operators were not identified in the advisory.

The attackers used publicly available AI tools to speed up several stages of an intrusion. CERT-In said Crafty Falcon surveyed targets with machine intelligence and automated break-ins “at machine speed.” This matters because faster discovery and exploitation can reduce the time defenders have to notice unusual activity or block access.

The key mechanism was the abuse of application programming interfaces, or APIs. These software links connect banks with non-bank lenders and fintech firms. In one reported police complaint, hackers compromised APIs and sent loan-disbursal instructions from two Mumbai lenders through accounts at ICICI Bank. More broadly, CERT-In said the group moved money using victims’ own payment and disbursement systems.

The attackers also selected holidays, weekends and other periods when monitoring was likely to be weaker. They returned to firms after incomplete remediation of earlier flaws. CERT-In called the threat active and ongoing, but did not connect specific reported incidents to Crafty Falcon.

The article describes a wider campaign spanning countries, organisations, money transfers and customer data. CrowdStrike said a China-linked group used ARTEX in an espionage campaign against more than 200 organisations in at least 14 countries. That figure concerns the broader ARTEX-linked campaign, not necessarily the Indian Crafty Falcon incidents.

In India, a police complaint said two Mumbai-based non-bank lenders lost about $2 million through more than 700 unauthorised instant transfers on October 8 and 9. The transfers came from accounts of ICICI Bank. Separately, State Bank of India disclosed a dark-web leak involving 3,897 customer records. Spandana Sphoorty Financial also reported unauthorised debits.

The article does not provide a total number of Indian victims, total stolen money, or total leaked records across all incidents. CERT-In did not name affected companies or say which incidents were linked to Crafty Falcon. AU Small Finance Bank was probing a potential customer-data leak.

When hackers compromise an API connecting a bank with a lender or fintech firm, they may be able to send instructions through a trusted software channel. The danger is that the receiving system can treat those instructions as part of a normal business process. A breach can therefore affect money movement, not just confidentiality or access.

The article gives a concrete example from Mumbai. A police complaint said hackers compromised APIs used by non-bank lenders and fintech firms to connect with banks. They sent loan-disbursal instructions, leading to more than 700 unauthorised instant transfers from ICICI Bank accounts. Two lenders reportedly lost about $2 million on October 8 and 9.

The incident shows why connected systems need careful monitoring and complete fixes. CERT-In said firms were attacked again after incomplete remediations. Financial institutions must detect unusual instructions, review API permissions and ensure that earlier vulnerabilities are fully closed, although the article does not detail specific controls.

An API, or application programming interface, is a software link that lets separate systems communicate and exchange instructions. In the article, APIs connect banks with non-bank lenders and fintech firms. They help those organisations work together, but they also create a route into financial processes that attackers may try to misuse.

The reported Mumbai incident shows the risk. Hackers compromised APIs and sent loan-disbursal instructions through connected systems. The complaint said this contributed to more than 700 unauthorised instant transfers from ICICI Bank accounts, costing two non-bank lenders about $2 million. The instructions used an existing payment connection rather than creating a wholly new route.

Careful control is essential because a trusted connection can amplify one weakness across several firms. Institutions need to watch what each API can do and investigate unusual instructions. The article specifically reports incomplete fixes and repeated attacks, but it does not list detailed API-control standards.