NEW DELHI: The Union Ministry of Electronics and Information Technology has selected Indian artificial intelligence company CoRover.ai to build a unified agentic AI platform for public services, with DigiLocker slated as the pilot deployment. Awarded by the National e-Governance Division through a quality-and-cost-based selection process, the contract will allow government departments to deploy task-oriented AI assistants on a shared infrastructure rather than building isolated systems. CoRover, known for developing BharatGPT, secured the contract with a perfect score of 100, ahead of Kyndryl Solutions, which scored 91.24. According to the company, more than 80 firms participated in the tender, with six ultimately empanelled for pan-India execution.
DigiLocker Chosen for Initial Four-Month Pilot
Under the initial rollout, DigiLocker will integrate an AI assistant to help citizens manage digital records and navigate connected government services. The digital document platform, which allows users to store and retrieve official papers electronically, will undergo a four-month development phase followed by 12 months of operations and maintenance support. This support period is intended to handle routine operations, track system health, and resolve technical issues during live deployment. The final scope of features available to the public will depend on which specific departmental databases are linked, as well as the underlying permissions and security controls governing citizen interactions.
Shared Infrastructure for Departmental Deployments
The planned platform aims to eliminate duplicated technical efforts across ministries by providing standard tools to create, deploy, and manage AI agents suited to distinct institutional requirements. Unlike conventional conversational chatbots, agentic AI systems are designed to complete multi-step workflows autonomously by interpreting instructions, gathering data, and interacting directly with integrated government software. The tender outlines operational tooling to supervise model behavior, maintain prompt repositories, and monitor response latency, task completion rates, execution failures, and computing costs. These metrics will allow system administrators to evaluate operational overhead, diagnose failures, and maintain systematic oversight over AI activity across public services.
Safeguards Target Jailbreaks and Prompt Injections
Given the sensitivity of citizen records, the tender mandates strict safeguards against emerging security vulnerabilities, including prompt injection attacks and unauthorized jailbreak attempts designed to bypass built-in system restrictions. AI agents interacting with public digital infrastructure will be restricted to authorized tasks, requiring verified user consent, routine authentication, and rigid system-level controls before initiating actions. Comprehensive monitoring of agent interactions and failure rates will be maintained to spot suspicious activity and rectify operational weaknesses, ensuring that document retrieval and automated service transactions remain strictly within approved security parameters.
Follow for daily updates on cybercrime, corporate fraud, DFIR, hacking, investigations, and digital forensics
