01.What was actually signed
Bengaluru-based AI startup Sarvam announced on 27 October 2026 that it has signed a five-year memorandum of understanding with Headquarters Integrated Defence Staff, the body that sits under the Ministry of Defence and coordinates the armed forces. The agreement gives Sarvam a single framework to design, develop and deploy AI applications for the Indian Army, the Indian Navy and the Indian Air Force, along with the joint and tri-service organisations that report to HQ IDS.
Under the MoU, Sarvam will build tools for operational needs, training and decision support, and will provide technical support after deployment. The company has not disclosed the financial value of the agreement, how many projects are covered, or when anything will go live. That absence of detail matters: an MoU is a framework to work under, not a purchase order, and it does not by itself confirm that systems are already in the field.
02.Inside Chanakya, the defence-facing arm
The projects will be executed by Chanakya, a firewalled division inside Sarvam that handles national security and defence work. It will draw on Sarvam's own AI stack, including its Sarvam 105B foundation model. Because all three services and the joint organisations will work off the same foundation models, agentic infrastructure, and geospatial and perception systems, what is learnt in a project for one service is meant to carry over to the others.
The technical scope includes fusing data from different military sensors so that personnel can assess a situation faster, and supporting target acquisition and the flow of information from sensors to the units that act on it. Sarvam says humans will retain control over operational decisions. The AI layer is to be built in India using Indian data and computing infrastructure, and deployed on infrastructure controlled within the country. Sarvam and HQ IDS also plan a wider AI layer that links different parts of India's defence capabilities.
03.Why defence AI is moving in-house
The agreement fits the government's Atmanirbharta push, which until now has mostly been about tanks, aircraft and electronics. Extending it to military AI means the software layer is treated as strategic infrastructure too. India has been funding domestic model building through the IndiaAI Mission, a programme of about 10,371 crore rupees that supports indigenous foundation model projects, and Sarvam is one of the companies backed under it.
The timing is not accidental. Governments elsewhere are signing military AI deals and running into questions about who controls the model. Sarvam co-founder Pratyush Kumar framed the deal in exactly those terms, saying the company exists to build AI in India, for India, and that the core of that mission is to put the country's own intelligence into the systems it depends on. Building on domestic compute and Indian data is as much a political requirement as a technical one.
04.What it means for students and job seekers
For engineering and computer science students, this is a signal about where AI hiring in India is heading. Defence work needs people who can handle data fusion, computer vision, geospatial systems, model evaluation and deployment on closed networks that cannot call out to a foreign cloud. Those are narrower skills than prompt writing, and they are harder to pick up from a short online course, which is why they command a premium.
The language angle is worth noting as well. Sarvam built its reputation on Indian-language AI, and multilingual tools that work in Hindi, Tamil or Bengali have obvious use in recruitment, training and public-facing systems. For a student choosing between a generic AI certificate and a specialisation in Indic language models, perception systems or AI security, this deal is a small pointer towards the second path.
05.What it means for startups and small businesses
For founders, the bigger takeaway is the procurement bar that is forming. Government AI in sensitive areas is moving towards Indian data, Indian compute and sovereign deployment, which shuts out tools that only run on a foreign cloud. Startups selling into defence, policing or critical infrastructure should assume they will be asked where their models are trained, where inference runs, and who can see the logs.
Small businesses outside defence will not be affected directly, but the skills and tooling built for this kind of work tend to flow outward. What remains open is accountability. Sarvam has promised human control over operational decisions, yet the agreement publishes no audit mechanism, no accuracy thresholds and no review body. Those details will decide whether military AI in India is genuinely supervised or merely described as supervised.
Frequently asked questions
What is the Sarvam defence MoU about?
It is a five-year agreement between Sarvam and Headquarters Integrated Defence Staff to build AI applications for the Army, Navy and Air Force. The work covers operational, training and decision-support tools, plus support after deployment.
How much is the deal worth?
Sarvam has not disclosed the value, the number of projects or deployment timelines. An MoU sets a framework for work rather than committing money to specific systems.
Will AI make decisions on its own in the military?
Sarvam says humans will retain control over operational decisions, with AI used for data fusion and decision support. The agreement does not publish details of how that oversight will be audited.