A desi AI kill chain? Indian military steps towards battlefield AI

India’s armed forces are preparing to put artificial intelligence (AI) closer to the heart of military operations, from analysing battlefield intelligence to helping identify targets and speed up decisions on strikes. These can include a desi AI-enabled kill chain, too.

A five-year agreement between Headquarters Integrated Defence Staff (HQ IDS) and Bengaluru-based Sarvam AI aims to develop a common AI framework for the Army, Navy and Air Force, including data fusion across the sensor-to-shooter loop.

Also Read: Sarvam enters defence AI with projects across Army, Navy and Air Force


The initiative marks a shift from isolated AI applications towards a connected system that could help commanders make sense of information flowing in from multiple military platforms.

Importantly, Sarvam says operational decisions will remain under human oversight, leaving no chance for an AI system to operate on its own during a conflict.

Why Sarvam AI?

Sarvam’s selection reflects a strategic requirement for control over the AI systems handling sensitive military data.

Military AI will handle sensitive intelligence, operational plans and potentially classified information. Building the AI layer in India and hosting it on sovereign infrastructure is intended to give the defence establishment greater control over its data and technology.

Unlike companies that primarily build applications on top of models developed elsewhere, Sarvam is developing its own foundation models and AI infrastructure in India. Its sovereign AI approach aligns with the military’s need to keep sensitive data, model deployment and critical technology infrastructure under Indian control.

Also Read: We have to develop our own AI layer for military services: Defence Secretary

Sarvam is not the only Indian company working on indigenous AI, however, and the MoU does not establish that it was selected because it is the only company capable of building foundation models. Its ability to offer models, agentic infrastructure, and geospatial and perception systems under a common framework may also have made it a suitable partner for the armed forces’ cross-service AI ambitions.

An AI layer across the three services

The MoU with Sarvam AI establishes an umbrella framework for AI projects across the three services and organisations under HQ IDS. Sarvam will design, develop and deploy applications for operational requirements, training and decision support, while providing technical support after deployment. Neither side has disclosed the agreement’s financial value, individual project details or deployment timelines.

The work will be handled by Chanakya, Sarvam’s dedicated, firewalled division for national security and defence. Its technology stack includes Sarvam 105B, the company’s large foundation model, alongside agentic infrastructure and geospatial and perception systems.

The ambition is to give all three services access to common AI infrastructure. Lessons from a project in one service could inform applications elsewhere instead of each force developing its own disconnected tools. Sarvam calls the proposed integration an “AI fabric” linking different elements of India’s defence capabilities.

A desi AI-enabled kill chain?

Sarvam’s announcement identifies decision support, target acquisition and data fusion across the sensor-to-shooter loop as key areas of work.

The loop describes how information moves from sensors that detect a potential threat to the systems and personnel responsible for assessing and engaging it. Satellites, radar, drones and other intelligence sources can generate enormous volumes of data. AI could help sift through those inputs, identify objects of interest, correlate observations and present commanders with a clearer operational picture.

It could help prioritise potential targets or provide information relevant to weapon selection and mission planning. One key objective of AI on the battlefield is to compress the kill chain, reducing the time between detecting a target and acting on it.

To be sure, these are possible applications of the proposed framework, not confirmed descriptions of systems to be built under it. “The focus areas are decision support, target acquisition and data fusion across the sensor-to-shooter loop, with human oversight retained on operational decisions,” Sarvam has said in the press release.

Also Read: AI an economic equaliser, tech at its best when inclusive, democratic: Google India's Preeti Lobana

Sarvam explicitly states that human oversight will be retained over operational decisions, which would mean an AI system could identify and recommend a target but would not be authorised to decide independently that it should be attacked.

Lessons from the US war on Iran

The US military’s use of AI during Operation Epic Fury, its campaign against Iran launched on February 28, 2026, offers a glimpse of how these systems could change targeting operations.

According to media reports, the US military used Palantir’s Maven Smart System, which incorporated Anthropic’s Claude, to analyse classified intelligence and help identify and prioritise targets. The system drew on information from satellites, surveillance and other intelligence sources. Reporting said the US struck more than 1,000 targets in the first 24 hours of the campaign.

Maven helped compress work that traditionally required analysts to review large amounts of information and prepare target lists. The system did not simply control weapons but also supported the intelligence and planning process that preceded military action.

The campaign also exposed the risks of relying too heavily on AI-assisted workflows. Reportedly, a Pentagon probe into the February 28 strike on a school in Iran found that flawed intelligence, outdated imagery and overreliance on AI contributed to the attack, which killed more than 150 people, including at least 123 children. The full investigation has not been publicly released, and the precise contribution of the AI system remains contested.

That is why Sarvam has emphasised human oversight and control of battlefield AI systems.

India has already used AI in operations

The Indian military is not starting from scratch with Sarvam. During Operation Sindoor in May 2025, the Army used AI for intelligence analysis, threat assessment, predictive modelling, weather forecasting and the creation of a common operational picture.

Lt Gen Rajiv Kumar Sahni, who was Director General of Information Systems during the operation, told The Times of India in October 2025 that AI was extensively used for real-time multi-sensor and multi-source data fusion. He said 23 applications with specific tasks were used to process data and inputs. AI-assisted analysis contributed to 94% accuracy in targeting Pakistani positions and terror infrastructure.

The Army has since showcased an AI-based satellite imagery analysis system that identifies changes and flags important observations. It has also demonstrated a portable AI system designed to support analysis and planning in areas without internet connectivity, alongside Ekam AI, an indigenous platform for handling information and supporting decisions in secure environments.

AI beyond the battlefield

AI work is also underway across the other services. The Navy’s 2024-25 annual report lists Antarbashi.AI, an offline language translator used on anti-piracy deployments, and Satyapan.AI, a facial-recognition system used to help identify suspected pirates.

Its NARAD platform is a private, offline language model for searching and summarising official naval documents. HawkSecure uses AI and machine learning to analyse network traffic and detect anomalies. Drishti.AI, designed to monitor personnel movement and detect man-overboard situations, has undergone prototype trials aboard INS Visakhapatnam. These examples range from operational deployments to trials rather than representing a uniform level of maturity.

The Army is also exploring autonomous systems. A Defence Ministry report says an autonomous weapon system experiment under Mission SWAYAT was conducted at Babina Field Firing Range in April 2025 to evaluate indigenous capabilities and develop a roadmap for possible induction. However, the experiment does not establish that autonomous lethal systems have been inducted into operational service.

The Sarvam agreement could help connect these separate initiatives under a common framework. Whether that becomes a genuine battlefield advantage will depend on how well the systems work together, how securely they handle sensitive information and whether their recommendations remain trustworthy in combat.

For now, the MoU is a framework for development, not evidence that India is developing a fully integrated AI kill chain. Its significance lies in the direction it sets: bringing sovereign AI into the process through which military intelligence becomes operational decisions, while keeping responsibility for those decisions with human commanders.