The future in the present: AI in the Indian military

  China has been pushing its military to adopt AI for almost a decade. The United States (US) has made complete AI integration before 2030 a part of its military strategy. 

  Where is India, which will host an international AI summit this month? 

  India’s armed forces are adopting AI, interviews suggest, but analysts urge advancing it. Or else, they say, the 

already significant technological gaps with the US and China will widen.

  Globally, AI was formally born during a US university conference in the 1950s. But a large language model (LLM) was created only in 2020. Commercial AI took off with ChatGPT two years later.

  The importance of AI in defence is set to grow, Indian Army chief General Upendra Dwivedi said at an annual news conference in January. To enhance multidomain combat readiness, the Army is undergoing changes, including moving towards an aviation brigade and a rocket-missile unit.  The Army’s modernisation priorities with AI are “knitting” its legacy equipment, improving the mobility and protection of troops and physical assets, strengthening network operations, data centricity and space and satellite communication, and acquiring more new-generation systems, including unmanned aerial vehicles (UAVs) and counter-UAV technology. To modernise logistics, robots and mule drones have been inducted in large numbers, but that is not enough, Dwivedi said. 

  “Moving ahead, we need AI practically everywhere, be it force visualisation or application or preservation.” 

  India’s armed forces are thought to have used AI’s predictive analysis of adversary behaviour and target identification during the four-day India-Pakistan conflict last year, when India struck multiple locations in Pakistan with clear precision. 

  “It is important for our kill chain, that from the sensor to the shooter, the connection is stable and timely. We have used this kill chain and because it was effective, we were able to address the target in the right place,” Dwivedi said. 

  The term “kill chain” is a military concept of attack. His remarks come amid a surge in the deployment of drones and other aerial items globally, in the contested air littoral, the low-altitude space between the Earth’s surface and a height of 10,000 feet, above which most piloted aircraft fly.  

In India, the vast majority of that space is filled with army assets, Dwivedi said, adding that software for “the management system of our flying equipment” will be created. In the future, the Army, the Indian Navy and the Indian Air Force (IAF) will be able to connect to the same systems, which is why AI integration is essential. 

  While India won the military conflict in May, media reports said Indian fighter jets were hit by Pakistan (and India hit Pakistan’s fighter jets). The IAF has not confirmed or denied that its jets were hit (only that “losses are part of war”) but said all IAF pilots had returned safely. 

  The hard lesson from Operation Sindoor is that advanced jets and missiles need wrap-around technologies, especially during bigger conflicts, military analyst Lieutenant General Raj Shukla (retired) said. 

  “A Rafale (fighter jet) with ‘fancy missiles’ is useless if you don’t have cyber (capability) to degrade the adversary’s command and control, or if you don’t have space to target it.”

  He cited Ukraine, which does not have a conventional air force or navy but has been able to deploy AI to sustain an asymmetrical war against Russia, a traditional military power, for four years. Ukraine’s Soviet-era defence-industrial base aside, the country had American technology entrepreneur Alex Karp build an AI-enabled software-driven command-and-control system. 

  “That has brought down decision-making time to nanoseconds and the kill chains have been compressed to minutes. They have developed these lethal frontiers where tanks, infantry, any movement is picked up and neutralised in 5-to-6 minutes, on a frontage of, say, 600-700 kilometres (km),” Shukla said.

  If India replicates that approach along its border with Pakistan and China, he said, “see what it will do”.  

Karp was on Time magazine’s list of the world’s 100 most influential people in 2025.

  India needs top technology talent, large language models and training, data centres, computing power and energy supplies to become AI-rich in defence, Shukla said.

  Ongoing work to publicly release India’s first national security strategy could offer more clarity and set expectations in the longer term for an AI strategy for defence, Antoine Levesques, senior fellow, International Institute for Strategic Studies, a London-based think-tank, wrote in 2024. 

  So far, though, no such strategy document is publicly available.   

  Application layer 

  India’s Defence Research and Development Organisation had made more than 75 AI products, ranging from autonomous robots and cyber defence to AI-based surveillance systems, by 2022. The government had then said ₹100 crore ($12 million) each would be given to the three services annually for AI support (some more funds have been announced since). China, by contrast, is estimated to spend at least $2-3 billion on military AI annually, and the US some $13 billion (figures vary). 

  Lieutenant General Harsh Chhibber, director-general, information systems, Indian Army, said in an interview with the Blueprint on January 23 that the Army’s current AI integration focuses on data digitisation, AI-enabled decision support, analytics and staff augmentation. AI is being employed to enhance situational awareness, reduce (human) cognitive pressure, and improve the speed and quality of decision-making under full human control.  

“A structured effort in ‘the (Army’s) year of networking and data centricity’ is under way to digitise brigades and higher formations by converting legacy, siloed and paper-based data into structured digital repositories. This includes operational, logistics, personnel, equipment and administrative data. This digitisation forms the foundational layer for meaningful AI application and data-driven command and control.”

  Among products, the Army has Ekam AI, an AI-as-a-service platform, to tie its use across operational and administrative domains, working as a “backbone” for applications such as AI chat, document-related work and specialised military bots. It focuses on automating routine staff function, improving consistency and accuracy of outputs, and enabling faster access to authorised information, while ensuring data sovereignty and compliance with the Army’s secure networks. Another is Sama Drishti, an AI-enabled tool that assists formation commanders and the rank in understanding real-time operational scenarios. 

  “It allows authorised users to interact with operational and intelligence data using natural language, akin to engaging with a digital staff officer. The system interprets ‘commander intent’, securely retrieves only role- and responsibility-specific data, and synthesises it into clear, operationally relevant insights, including narrative assessments and GIS (geographic information system)-based outputs,” Chhibber said.

  GIS helps the military analyse spatial data such as of terrains and from satellite imagery. 

Planned AI applications and platforms include: Ekam LLM, the first sovereign language-learning model in Indian defence, which is being developed; and agentic AI, an autonomous system that can meet complex objectives with minimal human intervention. In addition, some 100 ‘AI agents’ will be hosted on air-gapped (isolated) army-captive data networks. The Army might use AI for independent tactical or lethal decision-making later. 

  “The Army continues to operate under human-in-the-loop and human-on-the-loop principles, ensuring that command authority and accountability remain firmly with human commanders. Fully autonomous combat roles and delayering of decision-making is in progress,” he said.

  Human in the loop is when people improve the efficiency of a machine through active participation and human on the loop is people providing feedback to a machine to improve its performance.  

Large-scale predictive AI applications such as forecasting operational outcomes, logistics demand, equipment health or personnel trends are still in the development stage in the Army. Such capabilities depend on sustained availability of data, consistency and integration across formations and systems. Integrating existing weapons platforms and combat systems with AI-driven functions is also “in process”.

  “The integration of AI into weapons, particularly for targeting, engagement-support or fire-control requires extensive validation, high reliability under combat conditions and alignment with established doctrines. Consequently, such applications are being pursued cautiously and incrementally,” Chhibber said. 

  “The overall adoption of AI remains primarily at the application layer, with several important areas still under development or evaluation.”

  An image-and-video repository has been created for AI models with more than 200,000 images of army equipment. Shakti, another data project, when fully developed, will be a land vector command-control and communication system (digital battlefield management platform) that is “pivotal for sensor-shooter integration”. Data security is vulnerable. The Army is working on maintaining data integrity to ensure it remains consistent over time, “regardless of internal or external influences”. Users of the organisation’s data assets will rely on the same information for decision-making and analysis.  The Army is developing applications “with inherent security consideration” in mind. It has also set up more than 100 “auto weather stations” along the country’s northern borders that provide warning of natural disasters within 100 kilometres of the international border (with Pakistan) and the LAC. Doppler radars are being installed for long-range weather detection.  

A “canopy inspector” for fighter jets that can detect defects based on AI inputs (Photo: Ministry of Defence)

  The OODA loop 

  The IAF does not share information about the use of AI in operations. According to a source, its air-gapped network and predictive analysis were used effectively during Operation Sindoor. By many accounts, the IAF has been at the forefront of AI adoption among the three services (a global trend) — also because it has technology-heavy equipment.  

The IAF is refining its language-learning models, and uses AI in blockchain, digitisation, encryption, data certification, software systems and medical work, as well as to improve supply chains, including during conflicts, and to find alternative solutions. The source said AI shortens the OODA loop, short for “observe, orient, decide, act” — the military decision-making model proposed by a US Air Force officer in the 1970s.

  An IAF source with direct knowledge of AI adoption, said, “National security decisions cannot be left to machines, which is why we have the human-AI teaming approach, especially for operations.” 

  The AI-integration framework has been built upon decades of modernisation in the IAF. The foundational steps were taken over 2018-22, with the main objective of analysing and utilising a huge amount of data collected. 

  As India shifted its defence policy to self-reliance, the need to create an ‘informatised system’ became clear. The IAF’s AI cell was set up in 2019. To meet real-world operational challenges and to keep up with technological changes, Udaan or the Unit for  Digitisation, Automation, AI and Networking, was formed in 2021 (and the IAF’s AI Centre of Excellence in 2022). Udaan (“flight” in Hindi), where only IAF engineers work, specialises in computing, big data analysis and developing applications. The tasks include absorbing operational data across applications and automation integration. 

  “We have an AI suite and the natural language-processing capability (end-to-end),” the source said. “In terms of document analysis, we are not looking at AI to just summarise but for speech synopsis, say, when operational data is generated. All the data is transcribed and converted to text.”  

  Other AI use includes predictive maintenance of equipment (including radars and aeroengines), based on an analysis of historical data, or AI-enabled prior knowledge of repairs and service life; personnel management (who’s the right fit for what job and where); and logistics (where spares are located and the fastest way to get them).  

A major challenge is that the IAF has a mix of old (some more than 30 years) and new platforms and systems, which in turn means heterogenous data. For AI to function optimally, it has to make the correct data distinction. All datasets must meet the metadata standard. Plus, data-use requires creating standardised guidelines and frameworks that work within the air-gapped network. The IAF plans to acquire hardware and integrate it with the software developed internally for new AI applications later this year.

  The visual maintenance of aircraft is done by drones. Conversational AI is used to study air accidents, including those caused by bird hits. 

  Wing Commander Prateek Thapar (retired), who was the first officer in the IAF’s AI cell, said AI-enabled image intelligence is very important for precision strikes by missiles.  

  Intelligence analysis in general is a big part of AI activity. 

  “A plethora of information comes in every day, gleaning and linking the data that will be needed for operations later. It’s important to do that in peacetime,” Thapar said.

  AI improves situational awareness, but the decision to engage or not to engage is taken by humans, he said, adding that India will need an AI command at the level of battle centres.

  In 2005, the IAF’s internal network was established. Operations and the monitoring of operations were digitised after that. By 2025, AI had touched all aspects of the organisation, from the integration of weapons to operations, training, wargaming, maintenance, logistics, materials and the auxiliary system.

  The Navy is integrating AI and machine learning across critical mission areas to enhance operational effectiveness and to maintain maritime superiority, according to an Indian Navy statement to the Blueprint. The Navy has some 50 AI projects, covering language translation, target identification from imagery, sensor enhancement, inventory management, perimeter security, and maritime domain awareness. Future AI work will include LLM-based applications, offline language translators, autonomous vessels, smart ship technologies, and unifying enterprise data.

  “These initiatives aim to deliver tactical and strategic impacts by accelerating the OODA loop and enabling smarter devices for actionable intelligence from diverse data sources like imagery, machinery, and weapon systems.”

  Strategic impact-and-autonomy integration is reshaping naval doctrines elsewhere. The Indian Navy is exploring autonomous systems to remove humans from data-gathering and basic-decision loops, including large displacement underwater vehicles for undetected operations. 

  Developments in AI-driven weapons in advanced countries that analyse and engage targets autonomously are prompting doctrinal shifts in the Indian Navy, focusing on core elements like annotated datasets and data science, the statement said.

INS Valsura in Jamnagar, Gujarat, has housed the Navy’s Centre of Excellence in Big Data since 2020. An AI-training roadmap for officers and sailors is coming. 

Sarvatra Pehchaan, an AI-enabled system with indigenous software, to detect border breach (Photo: Ministry of Defence)

AI warfare

  Gregory M Reichberg, professor, Peace Research Institute Oslo, a think-tank, who has written on AI in Indian defence, said India has worked a lot on swarm warfare (developing applications). But it could be farther on AI, given the country’s software skills. 

  India is considered a top AI power, with the US and China leading the global race.

  There is an overreliance worldwide on autonomous weapons systems, to identify and engage targets, which may be more important for India, going ahead, Reichberg said. But AI, which is useful for battlefield coordination, should not substitute command decision-making, only support it. 

  He said an area India could look at is AI application awareness and interfacing, especially to move military assets effectively after locating the adversary’s position. 

  In his country, Norway, a population of about 5.65 million protects a landmass of about 385,207 square km, where the terrain is inhospitable in many parts. There, AI can provide inputs on which military assets are most suitable to fend off an adversary across domains, he said. 

  “There are exaggerated claims that AI can make decisions, but on tactical and strategic levels humans must be involved, so that pitfalls are avoided.”

  The Indian military will need to consider the development of adversarial AI in some other countries. Adversarial AI is that which manipulates machine data to trigger incorrect decisions.

  Machine-learning vulnerability is another concern in AI use. Although military systems are secure because they do not use open channels, even military communication links can be hacked into. “Unless there is a breakthrough in quantum (unbreakable codes are created), data-poisoning can happen within the model of language-learning itself,” Reichberg said.

  A mantra of military data-use is “harvest today, decrypt tomorrow”. 

  In 2023, a special team was formed in the IAF to study the changing quantum computing landscape. Post-quantum cryptography is being worked on to protect data. The Indian military establishment also maintains a high standard of clearing algorithms. But both funds for technology and policy focus are needed.

  A framework on integrating quantum technologies — communication, computing, sensing and metrology, and materials and devices — in the Indian military was announced on January 22.

  “While we are imbibing AI into the services, we need to go miles ahead,” Thapar said. 

  The Indian military has to move beyond the use of AI for information-gathering, analysis or the simulation of operations and applications to innovation, analysts interviewed said. 

  “In terms of valour and other combat matrix, ours is a very professional military. It is ironic that we are not lacing it with technology, in the pace and scale needed,” Shukla said. “Combat-AI, not (only) logistics.” 

There’s a craze for swarm drones in the world but they are low-to-medium weapons, Thapar said, adding that India should expand the inventory of AI in its arsenal with advanced implements. “Autonomous weapons can be the first line of defence. We need predictive analysis (air, naval and land), and a separate cyber command.”

  India has a responsible position on lethal autonomous weapon systems (LAWS), Thapar said, “but the adversary could deploy AI indiscriminately”, particularly because international regulation is still thin. 

  While the United Nations General Assembly adopted a resolution on LAWS in December 2023, the conclusion of a legally binding instrument, prohibiting systems that cannot be used in compliance with international humanitarian law, is pending.  

Reichberg said Israel made “indiscriminate use of AI” in Gaza (2023-25), not just with the targeting of locations but also the frequency and the spread of bombings.

  China’s “overinvolvement” with AI in its armed forces is a concern, Thapar said. “The Indian government needs to think about military AI.” 

  In a 2025 report to the US Congress on “military and security developments” in China, the US Department of Defense said, “China believes the next revolution in military affairs will occur when militaries transition to ‘intelligentized’ warfare and fully integrate AI, big data, advanced computing, and other technologies into the joint force.” 

  The People’s Liberation Army (PLA) is still developing the military theory, doctrine, and operational concepts for this warfare and it continued experimenting with different AI-enabled capabilities and concepts throughout 2024, it added. 

  China’s AI trajectory can pressure India into a defence-heavy posture. But India can strengthen deterrence without sacrificing growth (economic) and even position itself as “the world’s strongest dual-use AI democracy”, Thapar said. 

  AI itself is transforming, “there’s nothing futuristic about it”, Shukla said, adding that a missile coming in at Mach 7 or Mach 8 (seven-eight times the speed of sound), cannot be intercepted at “the rate of human judgment”.

  Revenue and capital figures in the defence budget do not indicate combat capability, he said: “How much are we spending on AI?” 

  India should work not just on AI, but edge-computing, robotics, 3D printing, and other new technologies that are all being woven into military capacities. The PLA set up its strategic support force in 2015 for AI-enablement of sensors and shooters across military domains, he said.

  “Our power differential with China is $450 billion year on year. With technological innovation and civil-military fusion, we can deter China asymmetrically”, he said.  

While indications are that technologically, China is doing very well, all military capabilities have to be combat-proven. “There’s a limit to winning without fighting.”

  On the other hand, US President Donald Trump’s proposed Golden Dome (air defence system against ballistic and hypersonic missiles) is expected to see a spending of $175 billion over the span of three years. The US is concerned about the PLA’s rocket force. India should be much more concerned, Shukla said. 

  India’s military curriculum must scale up. But the larger issue is the delivery of AI to fleets and formations. The era of software-enabled warfare is here, and AI is reinforcing it.