India will need to develop its own artificial intelligence models, train them on indigenous military data and accelerate defence technology procurement, if it is to keep pace with the rapidly changing nature of warfare, former DRDO Chairman Dr Samir V Kamat said, addressing the inaugural edition of the ‘CHAKRAVIEW’ symposium: ‘The Algorithmic Battlefield’ on August 6 at United Services Institution, New Delhi.

Describing AI as a fundamental shift in the character of warfare, Dr Kamat said military operations are moving from human scale reaction speeds to machine speed execution, with data, algorithms and autonomous systems increasingly influencing decisions on the battlefield. “For centuries, the fundamental constraint of warfare has been the human nervous system,” he said, explaining that the speed of battle was traditionally determined by how quickly a human could process a threat, decide on a response and issue orders. That same constraint is now being challenged by AI.

Dr Kamat said the traditional ‘Observe, Orient, Decide and Act’ or OODA cycle is being compressed as military systems increasingly connect sensors, computing platforms and weapons through digital networks. “We are transitioning from human scale reaction speeds to machine speed execution,” he said. “Victory will belong to the side that can process data, generate options and execute actions the fastest.”

According to Dr Kamat, the defining feature of algorithmic warfare is the increasing automation of the ‘sensor to shooter’ pipeline. Modern militaries are dealing with enormous volumes of information generated by satellites, unmanned aerial vehicles, drones, cyber sensors and radar networks. For human analysts, processing this information at battlefield speed is getting increasingly difficult. AI can bridge that gap by identifying patterns and potential targets across thousands of hours of surveillance footage, while also combining information from different sources.

Citing an example, Dr Kamat pointed to the US Department of Defense's ‘Project Maven’, in which Palantir played a significant role, as an early example of AI being used to accelerate the analysis of military intelligence. But the role of AI is moving beyond target recognition. Advanced systems can increasingly combine operational histories, weather information, logistics data and information about adversary behaviour to predict possible courses of action and recommend responses to commanders.

“The algorithm transforms the fog of war into a calculated matrix of probabilities,” Kamat said.

The growing role of AI also raises the question of how much authority should be given to machines in lethal operations. He outlined three broad levels of military autonomy: human in the loop, human on the loop and human out of the loop. In a human in the loop system, AI primarily acts as an analytical assistant, identifying potential targets while a human operator retains the final authority to engage. Human on the loop systems allow greater autonomy, with machines capable of operating independently within defined parameters while a human monitors the system and retains the ability to intervene. He said the most consequential would be the human out of the loop, owing to the autonomy it provides- where weapons can identify and engage targets without direct human intervention. While autonomous systems can offer significant tactical advantages, Kamat warned that their widespread deployment could create new strategic vulnerabilities.

One of the most serious risks what he described was the ‘Flash War Paradox’. If two highly automated military systems confront one another, decisions that previously took hours could be compressed into seconds. An algorithm could misinterpret an ambiguous movement as an attack, triggering an automated response that is then interpreted by the adversary as further escalation. “This could escalate into a full-scale kinetic conflict before a human president or prime minister even receives the first briefing,” he warned. Such systems could therefore compress the time available for political leaders and diplomats to intervene, creating what Kamat described as a potentially volatile and brittle defence ecosystem.

The other major concern is the ‘black box problem’, he said. Deep learning systems often identify statistical patterns without providing humans with a clear explanation of how they reached a particular conclusion. In a military environment, this creates serious consequences if an AI system incorrectly identifies a civilian object as a military target. Military AI systems are also vulnerable to adversarial attacks, where an opponent deliberately manipulates inputs to deceive a computer vision or machine learning system. “Software code can dictate life and death on a global scale,” Kamat said, stressing the need for responsible AI frameworks and international norms.

For India, the priority must be to develop indigenous AI capabilities rather than rely on large language models developed abroad. “We cannot depend on LLMs developed by foreign countries,” he said, stating that India will need to develop its own models and focus particularly on smaller language models, or SLMs, suited to specific military requirements. The models will also need to be trained on India's own data. “We have to train our models on our own data, not on foreign data or synthetic data,” Kamat said.

Military AI will also require a fundamentally different computing architecture from that commonly used in the civilian sector. He cautioned that conventional cloud-based AI systems could face serious limitations during a high-intensity conflict. Satellite communications and other communication links could be jammed, spoofed or severely constrained by bandwidth limitations.

The solution, he said, lies in expanding India's capabilities in edge computing, allowing AI systems to process information locally rather than relying entirely on remote cloud infrastructure. Dr Kamat said DRDO laboratories are already working on both the hardware and software required to develop next generation neuromorphic computing capabilities. Procurement must match technology. Even with technological capabilities available, Kamat identified defence procurement as one of the biggest obstacles to India's military AI ambitions. “The speed at which the technology is changing is now at least 10 times faster than what our procurement processes are,” he said. This mismatch, he argued, requires India to rethink how defence technology is acquired and integrated into the armed forces. The country will need to find ways to leverage technological capabilities scattered across its military, research institutions, industry and academic ecosystem while ensuring that procurement mechanisms can keep pace with rapid innovation.

The emerging AI battlefield will require a combination of indigenous data, sovereign AI models, resilient computing, faster procurement and responsible human oversight. For India, the challenge is no longer whether AI will become central to warfare, but whether its military technology ecosystem can evolve quickly enough to ensure that the country remains capable of designing, developing and controlling the systems that will shape the battlefield of the future.

Chakra Desk

Chakra News is a specialised digital media platform committed to in-depth coverage of India’s defence, national security, geopolitics, and critical technologies.