India's Ministry of Defence has signed a memorandum of understanding with the startup Sarvam AI to develop AI applications for the Army, Navy and Air Force, the company said on Tuesday. The agreement sets a framework for projects across all three services, focused on decision support, target acquisition and data fusion, and coordinated through Headquarters Integrated Defence Staff, the body that manages joint operations between the services.
Sarvam said the work will be carried out by Chanakya, a firewalled division it has set up specifically for national-security and defence work. Chanakya will apply the company's AI stack, including a foundation model it calls Sarvam 105B, along with AI agents and systems that process geospatial data and visual imagery. The stated aim is to let military personnel pull together information from many sources and surface the relevant pieces more quickly, with the same models and infrastructure shared across the three services so that lessons from one service's projects can inform the next.
The announcement is a framework rather than a fixed procurement. Sarvam has not disclosed a contract value, a timeline, or which specific systems will reach the field, and target acquisition and data fusion are capabilities that are far easier to name in a memorandum than to deliver and certify for operational use. What the deal does establish is a named domestic supplier for military AI, built on an Indian foundation model rather than a Western or Chinese one.
That sovereignty point is the reason the move matters. Sarvam was already positioned as a flagship of India's push to build its own large language models rather than depend on OpenAI, Google or Chinese labs, and extending that stack into defence keeps sensitive military data inside a domestically controlled system. It places India alongside a widening group of governments wiring commercial AI into their armed forces: Germany's Helsing unveiled an AI air-defence platform this week, and the United States has stood up a task force to coordinate how AI developers work with government. In each case, startups that began with consumer or enterprise products are now supplying the military.
The open questions are the ones a memorandum does not answer. Decision support and target acquisition put AI close to the use of force, and India has not said what human oversight will govern systems that help identify targets, nor how the models will be tested before deployment. Sarvam's defence division is new, and its track record in operational military settings is unproven. The agreement signals intent and a national preference for home-grown AI; whether Chanakya's tools work reliably under field conditions is a question that only deployment will answer.