Financial institutions in India are facing repeated cyberattacks from a newly documented threat actor called Crafty Falcon that deploys artificial intelligence and agentic AI tools to breach systems and steal funds, according to a security advisory issued November 4 by the Indian Computer Emergency Response Team (CERT-In). The agency warned that non-banking financial companies—including digital lending platforms, fintech firms, payment aggregators, payment service providers, and cryptocurrency exchanges—have fallen victim to successive attacks despite earlier warnings. The threat actor exploits application programming interfaces normally used for payment and disbursement transactions to execute unauthorized transfers, frequently timing attacks during holidays, weekends, and other periods when monitoring is reduced.
Crafty Falcon typically breaks in through internet-exposed non-production environments such as user acceptance testing, staging, and testing systems, then migrates into production environments by exploiting shared credentials, authentication tokens, or network connections between the environments. The attacker reverse-engineers application codebases to build custom exploits, then uses altered application packages to trigger unauthorized API calls that bypass security controls designed to trust only legitimate client applications. To evade IP allow-listing restrictions, the threat actor routes traffic through whitelisted third-party IP pools including content-acceleration services and app-store review systems—shared services that any internet user can instruct to fetch arbitrary URLs, allowing malicious traffic to originate from already-trusted IP addresses. Before executing high-value transfers, the attacker runs micro-value test transactions as small as one rupee to confirm that unauthorized fund flows are possible. The actor also targets cloud infrastructure including compute, serverless, and management services such as AWS Systems Manager and Lambda to execute commands and deploy malicious files that blend into normal cloud operations, and steals wallet keys from cryptocurrency exchanges to drain digital asset wallets.
CERT-In and its Finance branch observed that Crafty Falcon generates large volumes of custom scripts rapidly, with functionally identical variants refined in quick succession—a signature consistent with AI-generated code and iterative AI-assisted development. The agency noted the use and combination of several AI-powered offensive security tools against affected organizations, including penetration testing tool ARTEX based on PentAGI (an autonomous multi-agent framework that plugs into large language models), penetration testing framework CyberStrikeAI, and attack search engine Cairn. These tools enable attackers to "execute complete attack chains autonomously at machine speed, generate and refine large numbers of tailored attack scripts in rapid succession and adapt their payloads based on target responses," the advisory stated. While the tools are publicly available and could be used by many actors, most previous malicious campaigns deploying them were linked to China-based groups.
The attacks succeed partly because financial organizations performed incomplete remediation after receiving an earlier restricted-distribution advisory in early October about threat actors exploiting APIs for unauthorized transfers, CERT-In noted. The agency explains that the actor's ability to bypass conventional defenses stems from combining multiple advanced techniques: treating client applications as untrusted while exploiting server-side validation gaps, abusing cloud management tools that organizations assume only legitimate administrators use, and leveraging newly published vulnerabilities—including file-upload and deserialization exploit chains leading to web shell deployment in PHP and Node.js applications—as soon as they become public. The machine speed and adaptive payload refinement enabled by AI tools allow the attacker to outpace traditional detection and response timelines, especially during low-monitoring windows.
CERT-In provided extensive mitigation recommendations, including maintaining heightened 24/7 security monitoring with increased staffing and lower alert thresholds during holidays and weekends, centralizing real-time monitoring of cloud audit logs with alerts on high-risk events such as remote command execution and serverless function code updates, and separating non-production environments from production systems with no shared credentials, tokens, API keys, databases, or direct network paths. The agency urged organizations to extend real-time transaction monitoring to flag small-value test transactions to new beneficiaries followed by high-value transfers as strong indicators of potentially unauthorized fund movements, store cryptocurrency wallet keys in hardware security modules or cold custody, and prohibit unauthorized security and attack tools including publicly available AI-driven offensive frameworks on corporate infrastructure. Organizations should also monitor large language model API usage and keys, since LLM API traffic originating from servers or non-development hosts may signal attack infrastructure or a planted operator console, and adjust security information and event management correlation to identify machine-speed, multi-technique attacks with parallel threads, adaptive payload changes, and continuous activity outside business hours. The agency urged any organization detecting suspicious activity related to Crafty Falcon to report it immediately with all relevant logs. For financial sector leaders, the challenge isn't just adding defenses but rethinking the trust assumptions embedded in existing architectures—especially when adversaries can iterate and adapt faster than human security teams can respond.