In South Asia, Lethal Autonomous Weapon Systems (LAWS) have become a reality and not just a theoretical discussion. Loitering munitions and AI-powered surveillance systems have become part and parcel of Indian and Pakistani military strategies, and both were deployed in the May 2025 skirmishes along the Indo-Pakistani border.
However, the extent to which these systems are truly autonomous is disputed, they have yet to produce a weapon that identifies and kills without any human involvement, and the ultimate direction is clearly towards greater autonomy, quicker decision cycles and reduced human oversight.
With two nuclear powers, accustomed to crisis and thin margins for error, this move comes with risks far greater than on the battlefield; it is a compression of moments for prudence, it is a risk to the survivability of forces of retaliation, and it is a risk to the line between conventional skirmishes and strategic threats.
Having been one of the first and most persistent actors on the international stage that has advocated for international control on these weapons, Pakistan has the legitimacy and the obligation to influence the governance of the technology, both at the Geneva negotiating table and, more urgently, in its military doctrine and its relations with India.
But how autonomous are they? Not fully, yet. No state has deployed an autonomous weapon system that can, without human intervention, select and engage a target without any human decision-making in the targeting process itself; even the UN’s disarmament office admits there is no agreed definition of LAWS.
What does exist is a growing continuum: the “human-in-the-loop” systems requiring a human command to fire, “on-the-loop” systems that are supervised by a human, and “out-of-loop” designs that operate without real-time human supervision.
India has claimed that its doctrine is “human in the loop”, but its hardware is outpacing the guarantee: jet-powered swarms of loitering munitions capable of hitting targets 600 kilometres away and swarms of interceptors “that can respond at a much faster pace than any operator can keep up with”. The autonomy may be partial today, but the trajectory is not.
The concentration of these tools in South Asia is highly asymmetric. India maintains an Artificial Intelligence Military Council, a standalone Defense AI Project Agency, has deployed as many as 140 AI-enabled surveillance systems along the Pakistani border alone, is in the process of deploying ‘Nagastra’ loitering munitions and Israeli-origin ‘Harop’ and ‘Heron’ drones and has pledged to deploy as many as 30,000 border drones by 2027.
By contrast, Pakistan has counterbalanced with the Air Force’s Centre for Artificial Intelligence and Computing and the Army’s Cyber Command, with a greater reliance on Turkish and Chinese platforms. This asymmetry between the two arch rivals, along with an aggressively armed India, leads to hubris and miscalculation in times of war.
The risk is no longer theoretical. It has become an inevitable reality. May 2025 is a living example of this fact, and it demonstrated how drones enabled militaries to strike, and it demonstrated how drones enabled militaries to strike each other without calculating political costs or the threshold for escalation itself.
Given this, AI-powered sensors are more capable than ever before of finding the mobile missile platforms that give a more compact nuclear arsenal its survivability.
In addition to this, machine-speed decision cycles that take away the time for leaders to pause and confirm, and the logic of nuclear restraint, which was designed for hours and days, not seconds, start to break. This could even lead to a crisis if an Autonomous System Number (ASN) near the Line of Control gets hacked or broken, which neither country would have intended.
The most prominent shift that LAWs brought in warfare is that they separate precision from human decision-making. They allow states to project force and power without sending in people or political capital.
It encourages frequent, low-level incursions by drones that are gradually eroding deterrence norms established over decades, and it stealthily creates a gray zone between a skirmish and a strategic mission. The swarm that strikes a border post today may be used to target a nuclear-related target tomorrow.
Pakistan has always dealt with this debate responsibly. It was the first country to request a ban on autonomous weapons during the UN Human Rights Council in 2013, and led the review conference of the Convention on Certain Conventional Weapons (CCW) in 2016, which led to the establishment of the Group of Governmental Experts on LAWs.
Pakistan has since presented a working paper for a legally binding Protocol VI, and co-signed a call to launch formal negotiations in September 2025, which has been signed by 42 states. Pakistan needs to take the same measures at home too.
It needs to carry out bilateral confidence building measures, build a hotline to report drone incidents on an immediate basis and a clear national “red line” that no autonomous systems, no matter how genuine or competent, comes close the chain of command of nuclear weapons.
In a region like South Asia, algorithms that can one day become a shadow of the bomb need to be restrained from the start because it may not be retrievable afterwards.
Harsa Kakar
Ms. Harsa Kakar is working as a Research Assistant at Balochistan Think Tank Network (BTTN), at BUITEMS, Quetta, Pakistan. She is an MS International Relations Scholar at BUITEMS, Quetta, and a distinguished graduate of International Relations from the University of Balochistan. She specializes in AI, Global Politics, Diplomacy, Soft Power, and Conflict Resolution.
