Introduction
The deployment of artificial intelligence in military targeting systems is reshaping the conduct of armed conflict in ways that the architects of modern international humanitarian law could not have anticipated. From loitering munitions that autonomously select and engage targets within a pre-defined area, to AI-assisted systems that generate target lists and prioritise strikes, the boundary between human-controlled and autonomous lethal decision-making is being eroded in real operational environments. The legal challenges this creates are profound: international humanitarian law is built on the assumption that a human being, a commander or a soldier, makes targeting decisions and can therefore be held responsible for violations. When the decision to kill is delegated to an algorithm, the legal architecture of responsibility, accountability, and compliance with the principles of distinction, proportionality, and precaution strains toward breaking point.
This article examines the legal framework of IHL as applied to autonomous weapons systems (commonly referred to as Lethal Autonomous Weapons Systems or LAWS), the concept of meaningful human control as the normative touchstone for lawful autonomous targeting, the stalled progress of the Convention on Certain Conventional Weapons (CCW) discussions, the humanitarian case for a pre-emptive treaty, and India’s cautious but increasingly engaged position in these debates. It also draws instructive comparisons with the landmines and cluster munitions treaty models.
Legal Framework
International humanitarian law does not prohibit any weapon per se unless it meets certain criteria: inherent indiscriminateness (incapable of being directed at a specific military objective, or having indiscriminate effects), disproportionate effects on civilian populations, or the infliction of superfluous injury or unnecessary suffering beyond what military necessity requires. These criteria are codified in Articles 35, 48, 51, 52, 55, and 57 of Additional Protocol I to the Geneva Conventions. Article 36 of Additional Protocol I imposes a positive obligation on state parties to conduct a legal review of new weapons, means, and methods of warfare to assess compliance with IHL. This provision is the principal legal hook for the LAWS debate: states with Article 36 review processes (including the US, UK, Germany, and Australia) are required to assess AI-enabled weapons systems before deploying them, but the methodologies for such reviews are not standardised and their outputs are generally classified.
The principles of distinction and proportionality are particularly challenging to operationalise in autonomous targeting. Distinction requires a targeting system to identify and differentiate between combatants or military objectives on the one hand, and civilians or civilian objects on the other. Computer vision systems trained on image recognition datasets have achieved remarkable accuracy in controlled conditions, but warfare presents adversarial environments in which combatants deliberately conceal themselves, use civilian objects, and exploit the system’s limitations. The proportionality assessment (whether anticipated civilian casualties are excessive in relation to anticipated concrete and direct military advantage) is inherently a value-laden judgment requiring contextual understanding and the weighing of incommensurable values, capabilities that current AI systems do not possess in any legally meaningful sense.
Developments at the CCW and Beyond
The CCW, a framework convention concluded in 1980 under UN auspices, allows protocols addressing specific categories of weapons. The CCW’s Group of Governmental Experts (GGE) on LAWS was established in 2014 and has met repeatedly, producing a series of “Guiding Principles” in 2019 that represent the only multilateral consensus on LAWS to date. These principles affirm that IHL continues to apply to LAWS; that human responsibility for the use of force must be retained; that a human-machine interaction is required in any use of lethal force; and that states should conduct legal reviews of new weapons. The principles fall well short of a prohibition or even a definition of LAWS, reflecting the impossibility of reaching consensus among states with sharply divergent interests.
The GGE’s progress has been repeatedly blocked by the major military powers. The United States, Russia, United Kingdom, France, Israel, and Australia have all resisted any binding instrument or definition that would constrain their development and deployment of autonomous systems. These states argue that existing IHL is sufficient, that a definition of LAWS is impossible to agree upon, and that any prohibition would disadvantage their militaries without preventing adversaries from developing the same capabilities. This dynamic has rendered the CCW process essentially gridlocked since 2019.
In contrast, a growing coalition of states, including Austria, Belgium, Mexico, Chile, New Zealand, and the majority of African and Latin American states, have called for a binding treaty prohibiting fully autonomous weapons and requiring meaningful human control over the use of lethal force. The ICRC, which served as the intellectual engine of the Ottawa Process for landmines, has issued a position paper (2021) calling for new legally binding rules that would prohibit autonomous weapons systems that are incapable of being used in compliance with IHL, and require human control sufficient to allow compliance in all other cases. Civil society campaigns, including the Campaign to Stop Killer Robots, have mobilised significant public opinion.
The operational reality on the ground has moved faster than the diplomatic debate. Ukraine has deployed domestically developed AI-assisted drone systems; Israel’s use of AI-generated target lists in Gaza (the “Gospel” and “Lavender” systems, as described in investigative reporting by +972 Magazine in 2024) has been the subject of intense international controversy, with critics arguing that the volume of AI-generated targets effectively automated lethal decision-making in a way that violated IHL principles; and Russia has deployed loitering munitions with autonomous terminal guidance against Ukrainian targets. The legal status of these deployments under existing IHL remains disputed, but they illustrate that the debate has already moved from theoretical to urgently practical.
Contemporary Issues and Analysis
The concept of “meaningful human control” (MHC) has emerged as the central normative organising principle in the LAWS debate. Championed by the ICRC and states including Germany and Austria, MHC requires that a human being exercise sufficient understanding and control over the lethal decision to bear legal and moral responsibility for it. The concept resists precise definition: clearly, a soldier who pulls a trigger after independent assessment exercises MHC; clearly, a system that autonomously identifies, selects, and engages targets without human review does not; the vast middle ground, covering everything from AI-generated target lists reviewed in thirty seconds to semi-autonomous systems with human authorisation at some stage of the engagement sequence, generates ongoing controversy.
The accountability gap created by autonomous targeting is legally distinct from the challenge of proportionality assessment. Criminal responsibility in IHL requires a person who gave an order, failed to prevent a violation, or directly participated in an unlawful act. If a LAWS commits an IHL violation (for example, kills civilians because its target recognition algorithm misclassified them as combatants), who bears criminal responsibility? The programmer who wrote the algorithm, the commander who deployed the system, or the state that approved it? None of these satisfactorily maps onto the traditional commander responsibility framework. This “responsibility gap” has been identified as a fundamental objection to fully autonomous weapons independent of their compliance with specific IHL rules.
The gender dimension of LAWS is underappreciated but significant: research suggests that AI systems trained on historical warfare data may perpetuate biases in target recognition, and that male combatant-coded appearance may be systematically confused with civilian appearance differently for men and women. The ICRC has flagged this concern, noting that IHL violations are not only quantitative (too many civilian deaths) but also qualitative, involving the dignity and individuality of victims.
Comparative and International Perspective
The Ottawa Treaty (Convention on the Prohibition of the Use, Stockpiling, Production and Transfer of Anti-Personnel Mines and on Their Destruction, 1997) and the Convention on Cluster Munitions (2008) both demonstrated that meaningful arms control treaties can be achieved outside the CCW framework and without the participation of major military powers. Both were negotiated through “Oslo Process” type coalitions of willing states and civil society actors, entered into force relatively rapidly, and created powerful normative dynamics that constrained even non-party states’ conduct. The United States never joined the Ottawa Treaty but has largely ceased using anti-personnel landmines except in the Korean Peninsula; the stigmatisation effect of the treaty produced de facto compliance pressures.
This model is the explicit inspiration for those advocating a LAWS prohibition treaty negotiated outside the CCW. The argument is that waiting for consensus in the CCW, where any single state can block progress, means indefinitely deferring an instrument that is urgently needed. The counterargument is that a treaty without the US, Russia, China, Israel, and the UK would lack the military significance to change operational practice.
The Nuclear Non-Proliferation Treaty model, in which major powers retain capabilities while imposing constraints on others, is also cited in LAWS debates, though critics note that nuclear weapons’ existential risk made the NPT’s political dynamics unique.
India’s Position
India has participated in CCW GGE discussions without committing to either a binding prohibition or an affirmation that existing IHL is entirely adequate. India’s formal statements have supported the continuation of the GGE process, called for a definition of LAWS, and emphasised the importance of human control over targeting decisions without specifying what level of control is required. India has neither joined the states calling for a prohibition treaty nor aligned with the US-led bloc resisting any binding outcome.
India’s domestic LAWS development programme is understood to be in advanced stages, with the Defence Research and Development Organisation (DRDO) and private sector partners developing AI-enabled drone swarms, loitering munitions, and autonomous underwater vehicles. India’s defence procurement policy as of 2024 includes “positive indigenisation” requirements for defence items, creating strong incentives for domestic LAWS development. This strategic reality shapes India’s international posture: it is unlikely to endorse a prohibition that would constrain Indian defence R&D, but it recognises the international legitimacy cost of being seen to resist all LAWS regulation.
India’s position at CCW GGE meetings has also been influenced by its experience as one of the world’s leading contributors to UN peacekeeping operations: Indian peacekeeping troops could potentially face adversarial LAWS in conflict zones, creating a self-interest in ensuring that LAWS deployments comply with IHL and that liability for LAWS-caused casualties is clearly attributed.
Practical and Policy Implications
The failure to reach agreement on LAWS regulation creates practical risks that extend beyond the philosophical accountability gap. As AI-enabled targeting becomes widespread, the risk of rapid escalation between adversaries both using autonomous systems increases significantly. Unlike human combatants who can exercise judgment about de-escalation, autonomous systems following pre-programmed rules may engage in sequences of action and retaliation at machine speed, leaving no space for human intervention. This escalation risk has been flagged in nuclear security contexts (where AI-enabled early warning systems that autonomously authorise nuclear launches represent perhaps the most catastrophic application of autonomous weapons doctrine) but applies to conventional conflict as well.
The insurance and compensation frameworks of IHL (obligations to provide medical care, pay compensation for civilian casualties, facilitate prisoner of war protections) are also poorly adapted to situations where a state deploys a LAWS that causes civilian casualties without any human having specifically authorised the precise lethal act.
Suggestions and Reforms
The most constructive immediate reform would be the adoption at CCW level of a definition of LAWS that establishes a clear category of fully autonomous weapons (those incapable in principle of complying with IHL because they cannot distinguish combatants from civilians in dynamic environments). Such weapons should be prohibited through a CCW protocol or, if CCW consensus is impossible, through an Oslo Process treaty. A second tier of regulation should establish minimum standards for human control in semi-autonomous systems, including mandatory legal reviews (under Article 36) with published methodologies and outcomes.
India could exercise meaningful leadership in this space by proposing a middle-ground definition within the CCW GGE that focuses on the human control threshold rather than on technological characteristics, thereby creating space for consensus among states with divergent positions on automation per se. India’s credibility as a major developing-country military power, its peacekeeping contributions, and its non-alignment tradition position it well to broker such a compromise.
Conclusion
The governance of autonomous weapons systems is one of the most urgent and most neglected challenges in contemporary international law. The gap between the speed of technological deployment and the pace of legal regulation has never been wider. IHL’s foundational principles are not technically incapable of application to LAWS, but their meaningful implementation requires the kind of human judgment and contextual appreciation that current AI systems cannot replicate. The resulting accountability gap, the prospect of lethal force being exercised without any human bearing genuine responsibility for its consequences, is legally intolerable and morally objectionable. The window for pre-emptive regulation, establishing norms before the technology is ubiquitously deployed, is narrowing. The international community, including India, must act with greater urgency than the CCW’s incremental process has so far allowed.
