“AI can inform decisions. Only humans can make the judgment and take responsibility,” Lt. Gen. Vipul Singhal said during a defence-centred session at the India AI Impact Summit, framing a debate about the role of artificial intelligence in India’s armed forces and the limits of machine authority. According to the summit coverage, senior officers, defence scientists, industry figures and academics all stressed that speedier analysis must not dilute command accountability.

Speakers converged on a single aim: use AI to amplify military capability while preserving moral and operational control. Industry and military panellists repeatedly warned that algorithms should aid rather than replace human judgement, and that accountability for lethal decisions must remain with commanders.

Army leaders described AI as already moving beyond experimentation into operational use. “AI is totally transforming the way we analyse, decide and act, and transforming warfare,” Brig. Deepak Kumar said, while Lt. Gen. Rajiv Kumar Sahni argued that combat effectiveness increasingly depends on engineering, sustainment and the velocity of decision-making rather than on platforms alone.

Sahni set out three areas where the service is seeking industry and academic partnerships: tighter sensor-to-shooter integration to shorten the time from detection to engagement; predictive tools to regenerate forces under resource constraints; and indigenous development of drones with robust navigation, control analytics and adversarial testing. The emphasis is on embedding analytics into existing systems rather than buying whole new platforms.

That logic underpins a deliberate “smartisation” strategy for legacy equipment. Maj. Gen. Mohit Gandhi argued that existing arsenals need not be discarded: retrofitting sensors, condition-monitoring and AI-driven maintenance can improve readiness, reduce human error and extend platform life, though he warned of shortages in labelled datasets and the need for explainable, jamming-resistant models that can operate on secure or offline networks.

Predictive maintenance emerged as a clear, non-contentious battlefield advantage. Maj. Gen. P. S. Bindra said vehicles and other systems are now “speaking to us” via sensors and data loggers; moving from calendar-based servicing to condition-based monitoring and models that estimate residual useful life promises higher availability in extreme operating environments. The Army plans indigenous R&D, pilot-to-scale procurement through GeM and wider deployment if trials succeed.

Ethical and legal concerns sharply defined a line that speakers insisted must not be crossed. Singhal recounted an incident in which a machine-recommended strike was halted because the commander recognised a civilian evacuation the algorithm missed: “The commander paused. What does the machine not know?” Panellists warned that compressed decision cycles increase escalation risk if commanders rely on opaque systems.

Several experts highlighted governance shortfalls unique to military AI. Commentators pointed to the problems of bias detection, model drift as battlefield conditions change, bounded operational envelopes, lifecycle controls including decommissioning, and oversight for agentic systems that can act with limited human direction. One contributor urged embedding responsibility across the AI lifecycle and retaining the ability to disable systems if control is lost.

Speakers also stressed strategic sovereignty: reliance on imported GPUs and cloud-based algorithms is a vulnerability in conflict, they said, arguing for sovereign compute platforms, edge systems and a dedicated defence AI mission. The Army’s move to integrate AI with existing command-and-control capabilities mirrors broader efforts such as indigenous systems that automate planning and targeting, while domestic counter-drone and wide-area defence projects illustrate the kind of homegrown platforms the service seeks. Proposals included a sovereign defence AI platform, specialised training for commanders and a joint military AI command, with an acceptance that iterative experimentation and occasional failure will be part of building resilient capabilities.

Source Reference Map

Inspired by headline at: [1]

Sources by paragraph:

Source: Noah Wire Services