The new AI Cybersecurity and Assurance Lab will focus on adversarial testing, agent containment, AI assurance and workforce development as financial institutions expand autonomous AI deployments.
The growing use of artificial intelligence in banking, payments, lending and insurance is creating a new cybersecurity challenge: how to secure systems that can access sensitive information, make decisions and execute actions with limited human intervention. Addressing this emerging risk, the Indian Institute of Technology Kanpur (IIT Kanpur) and Bharat1.ai have launched AI Kavach, an AI Cybersecurity and Assurance Lab dedicated to protecting AI systems across the financial sector.
Established through IIT Kanpur’s Wadhwani School of AI and Intelligent Systems (WSAIS) and Bharat1.ai (Intellineuro Private Limited), the lab will focus on testing the resilience of financial AI applications against adversarial attacks and identifying vulnerabilities before they affect production environments.
“AI security cannot be an afterthought when autonomous systems are accessing sensitive data, executing transactions and influencing critical financial decisions. It must be built into the architecture from the outset.”
— Umakant Soni, Co-founder, Bharat1.ai
Its research agenda includes adversarial red-teaming of financial AI systems, containment and sandboxing technologies for autonomous agents, and work on AI assurance, traceability and responsible vulnerability disclosure. The initiative will also develop specialised cybersecurity skills through fellowships, security exercises and industry programmes, targeting more than 300 students and professionals annually. Compute resources supported by the IndiaAI Mission will contribute to the lab’s research activities.
The initiative comes as financial institutions increasingly integrate AI into customer services, fraud detection, credit assessment and transaction processing. While autonomous agents can improve efficiency and decision-making, they also introduce risks involving unauthorised actions, data exposure, manipulation and unpredictable behaviour under adversarial conditions.
Umakant Soni, Co-founder of Bharat1.ai, said AI security must be embedded in system architecture as agents take on greater responsibility within financial operations. He said AI Kavach aims to build sovereign capabilities that allow India’s financial ecosystem to adopt AI with greater confidence.
IIT Kanpur will contribute academic leadership, faculty expertise and laboratory infrastructure, while Bharat1.ai will bring AI infrastructure, secure testing environments and pathways to translate research into practical applications.
Prof. Manindra Agrawal, Director of IIT Kanpur, said the collaboration reflects the institute’s focus on frontier AI research and the development of resilient technologies.
Beyond research, AI Kavach is intended to become a collaborative platform connecting financial institutions, fintech companies, AI laboratories, regulators and cybersecurity experts. Its objective is to develop shared evaluation frameworks and best practices for secure AI deployment.
The launch signals a shift in the financial sector’s cybersecurity priorities. As AI systems move from advisory roles to autonomous execution, conventional application and infrastructure security will need to be complemented by testing the behaviour, permissions and decision-making boundaries of AI itself. AI Kavach aims to help build the research and assurance capabilities required for that transition.
