As AI becomes deeply embedded across banking, digital payments, lending, fraud detection, customer service, compliance, and identity verification, AI Kavach will provide a collaborative platform to rigorously evaluate, stress-test, and secure AI systems before they are deployed at scale. The initiative brings together IIT Kanpur's world-class research expertise with Bharat1.ai's AI infrastructure and deployment capabilities to help build a secure foundation for India's next generation of AI-enabled financial services.
Artificial intelligence is rapidly evolving from an assistive technology to an autonomous decision-making capability. AI agents are increasingly able to access enterprise systems, interact with external services, retrieve sensitive information, and execute tasks with minimal human intervention.
Recent global AI security incidents, including a widely reported AI security testing event involving OpenAI and Hugging Face, have highlighted that even advanced AI systems can behave in unexpected ways, including escaping outside controlled environments. These developments underscore the growing need for robust AI security testing, containment, traceability, and assurance before AI systems are deployed in critical sectors. Further, current financial systems face advanced security challenges as frontier AI models reduce the cost and talent needed to launch sophisticated cyberattacks.
For India, where digital public infrastructure and financial technologies serve hundreds of millions of users, ensuring the security and reliability of financial systems has become a strategic imperative.
AI Kavach has been established to help identify vulnerabilities before adversaries do and to enable financial services companies to take preventive steps against such breaches.
Conduct controlled adversarial testing of AI systems deployed across banking, payments, lending, insurance, wealth management, and financial services to proactively identify vulnerabilities, model failures, prompt injection risks, agentic exploits, and emerging cyber threats.
Develop technologies for securely sandboxing AI agents, restricting system access, managing credentials, monitoring autonomous behaviour, implementing runtime safeguards, and enabling rapid containment of compromised AI systems.
Advance research on AI assurance methodologies, incident attribution, explainability, auditability, and responsible vulnerability disclosure frameworks to improve transparency, accountability, safety and trust in AI-enabled financial systems.
Build India's AI security workforce through specialised education, research fellowships, hands-on security exercises, and industry collaborations. AI Kavach aims to train more than 300 students and professionals annually, leveraging advanced AI compute infrastructure alongside cloud resources supported through the IndiaAI Mission.
AI Kavach is envisioned as a global collaborative platform bringing together academia, industry, financial institutions, fintech companies, AI research labs, cybersecurity experts, regulators, AI startups, and technology providers.
IIT Kanpur, through WSAIS, will provide academic leadership, frontier AI research, faculty expertise, and advanced laboratory infrastructure.
Bharat1.ai will contribute and enable globally benchmarked physical and AI infrastructure, secure testing environments, program management, deployment platforms, ecosystem partnerships, and translational pathways that enable research to move into real-world implementation.
The initiative aims to establish common methodologies, evaluation frameworks, and build a global repository of best practices that strengthen the security of AI deployments across India's financial sector while contributing to global research on trustworthy and safe AI.
