Evaluating the Adequacy of Existing International Humanitarian Law on Autonomous Weapon Systems (AWS)
A Doctrine Analysis of the Distinction,Proportionality and Accountability Paradigms in the Regulation of Lethal Autonomy
Author: Suchika, Chitkara University, Rajpura Punjab
ABSTRACT
The proliferation of Autonomous Weapon Systems (AWS) marks one of the most consequential developments in the conduct of hostilities since the advent of nuclear weapons, displacing the human finger on the trigger with probabilistic, machine-generated targeting decisions. This paper interrogates whether the existing corpus of International Humanitarian Law (IHL) anchored in the 1949 Geneva Conventions, the 1977 Additional Protocols, and customary rules of distinction, proportionality, precaution, and military necessity remains doctrinally adequate to govern weapon systems capable of independently selecting and engaging targets without contemporaneous human judgment. Employing a doctrinal legal methodology, the paper examines the applicability of Article 36 weapons-review obligations, the jurisprudential legacy of the International Court of Justice’s Nuclear Weapons Advisory Opinion, and the command-responsibility framework developed by the ad hoc international criminal tribunals, to test whether AWS can be meaningfully reconciled with a legal architecture premised on human agency, and reviewable judgment. The paper concludes by proposing a new legal paradigm centred on the principle of Meaningful Human Control (MHC), a tiered international instrument modelled on the Convention on Certain Conventional Weapons, and a rebuttable presumption of State responsibility for autonomy-attributable violations, as the most viable route to reconciling technological sovereignty with humanitarian protection.
Keywords: Autonomous Weapon Systems, International Humanitarian Law, Article 36 weapons-review obligations,Meaningful Human Control (MHC)
1. TO THE POINT
Throughout history, has been a realm where human judgment has been crucial, in the face of extreme moral dilemmas. International Humanitarian Law (IHL), which governs armed conflicts, was built on the foundation that a human agent is essential in making decisions about life and death capable of discretion, adapting to new information, and bearing legal responsibility for mistakes. However, the rise of Autonomous Weapon Systems (AWS) challenges this fundamental premise.The urgency of these issue evident nations has already deployed autonomous systems with increasing levels of decision-making autonomy in warfare. The United Nations Secretary-General has called for a binding treaty to ban AWS that operate without human oversight.
2. USE OF LEGAL JARGON
IHL does not regulate weapons categorically; it regulates conduct. The relevant question is not whether autonomy is inherently unlawful, but whether a given AWS can be employed, and is in fact employed, in conformity with the cardinal principles codified in the 1949 Geneva Conventions, Additional Protocol I of 1977 (AP I), and customary international law as restated by the International Committee of the Red Cross (ICRC) Customary IHL Study.
2.1 The Principle of Distinction and proportionality
Article 48 AP I require parties to a conflict to distinguish at all times between the civilian population and combatants, and between civilian objects and military objectives, directing operations only against the latter. Article 51(4)(b) AP I further prohibits methods or means of combat which cannot be directed at a specific military objective. The doctrinal difficulty posed by AWS is not that autonomy is inherently indiscriminate, but that distinction as classically conceived presupposes a contextual, qualitative judgment assessing hors de combat status, feigned surrender, or civilian character in ambiguous circumstances that current sensor-and-classifier architectures perform through pattern recognition rather than normative reasoning. Article 51(5)(b) AP I prohibits attacks expected to cause incidental civilian loss excessive in relation to the concrete and direct military advantage anticipated. Proportionality is widely regarded as the most cognitively demanding rule in the targeting cycle because it requires the balancing of qualitatively incommensurable values human life against military advantage informed by battlefield context that shifts in real time. The delegation of this balancing exercise to an autonomous system rises the question whether a machine can perform an evaluative judgment that the drafters of AP I understood to be irreducibly human.
3. THE ACCOUNTABILITY GAP: RESPONSIBILITY WITHOUT A RESPONSIBLE AGENT
IHL’s enforcement architecture both the State-responsibility regime under the International Law Commission’s Articles on State Responsibility and the individual criminal-responsibility regime under the Rome Statute of the International Criminal Court was built around a human decision-maker whose intent, knowledge, or negligence can be reconstructed after the fact. AWS strain this architecture along three distinct fault lines.
3.1 The Individual Criminal Responsibility Gap , The State Responsibility Gap
Article 30 of the Rome Statute requires that a crime be committed with ‘intent and knowledge, while command responsibility under Article 28 requires that a commander knew or should have known that forces under their effective control were committing or about to commit crimes, and failed to take necessary and reasonable measures. Where an AWS autonomously selects and engages an unlawful target through an emergent classification error unforeseeable to its operator or commander, neither the operator (who exercised no contemporaneous decision) nor the commander (who lacked the requisite knowledge) may satisfy the mental-element threshold, and the programmer or manufacturer typically falls outside the personal jurisdiction of international criminal law altogether.
State responsibility for internationally wrongful acts does not require fault in the same sense as individual criminal liability, but it does require that conduct be attributable to the State and that a breach of an international obligation be established. Attribution is comparatively straightforward where an AWS is operated by State armed forces; the harder question is establishing breach where the AWS’s conduct diverges from what any human commander could reasonably have foreseen given the ‘black box’ opacity of modern machine-learning targeting algorithms a phenomenon this paper terms the evidentiary opacity problem, distinct from, but compounding, the doctrinal responsibility gap.
4. CASE LAWS
4.1 Legality of the Threat or Use of Nuclear Weapons (ICJ, Advisory Opinion, 1996)
The International Court of Justice’s advisory opinion on nuclear weapons remains the single most cited authority in AWS scholarship, not because it addressed autonomy, but because it articulated the two ‘cardinal principles’ of IHLdistinction and the prohibition on unnecessary suffering as constraints applicable to any weapon regardless of whether it is expressly named in a treaty, and confirmed the continuing vitality of the Martens Clause as a mechanism for addressing rapidly evolving military technology. The Court’s refusal to declare nuclear weapons categorically unlawful in all circumstances, while affirming that their use must nonetheless conform to IHL, provides the structural template this paper adopts for AWS: a technology may be lawful in the abstract while remaining unlawful, or legally uncertain, in its concrete application.
4.2 Prosecutor v. Tihomir Blaskic (ICTY, Appeals Chamber, 2004)
The International Criminal Tribunal for the former Yugoslavia’s Blaskic jurisprudence refined the command-responsibility standard under customary international law, clarifying the ‘knew or had reason to know’ threshold for superior liability. The judgment is instructive for AWS because it demonstrates how exacting the evidentiary threshold for superior knowledge already is in relation to human subordinates; that threshold becomes correspondingly harder, and in some scenarios practically impossible, to satisfy where the acting ‘subordinate’ is a software system whose decision pathway is not fully interpretable even to its own operators.
4.3 Article 36 Review Practice and the CCW Group of Governmental Experts
While not case law in the strict sense, the accumulated practice of the CCW Group of Governmental Experts on Lethal Autonomous Weapons Systems functions as a quasi-jurisprudential body of interpretive guidance. Its 2019 Guiding Principles affirmed, inter alia, that IHL applies fully to AWS, that human responsibility must be retained for decisions on the use of weapons systems, and that accountability cannot be transferred to a machine.
5. ADEQUACY ASSESSMENT AND A NEW LEGAL PARADIGM
5.1 The Adequacy Deficit, Restated
The foregoing analysis does not support the claim that existing IHL is wholly inapplicable to AWS; the ICRC and the overwhelming majority of States agree that IHL applies in full. The more precise finding is that existing IHL is under-determinate at the point of application: it supplies the substantive standards (distinction, proportionality, precaution) but not the institutional machinery necessary to verify compliance ex ante (transparent, harmonised, and internationally reviewable weapons reviews), to attribute responsibility ex post where machine behaviour is not fully explicable (a revised standard of superior knowledge calibrated to algorithmic opacity), or to reach the full range of actors implicated in an AWS’s design and deployment (algorithm developers and supply-chain participants).
5.2 Meaningful Human Control as the Organising Standard
This paper argues that the most doctrinally coherent response is to elevate Meaningful Human Control (MHC) from a policy slogan to a binding legal standard operationalised through three cumulative criteria:
(i) Predictability – the operator must be able to reliably anticipate the system’s behaviour within its intended operating envelope;
(ii) traceability- the system’s targeting decisions must generate an auditable record sufficient to reconstruct the basis for an engagement after the fact; and
Rather than a categorical ban, which is unlikely to attract the ratification of major military powers, this paper proposes a tiered instrument modelled structurally on the CCW’s existing Protocol architecture: Tier I prohibitions for AWS incapable, by design, of satisfying the distinction principle in any realistic operating environment (a category-based prohibition analogous to the treatment of blinding laser weapons under CCW Protocol IV); Tier II conditional authorisation for AWS satisfying the MHC criteria, subject to mandatory, internationally registered Article 36 reviews with a standardised disclosure template.
6. CONCLUSION
Autonomous Weapon Systems do not exist in a legal vacuum, and the claim, sometimes advanced in public discourse, that AWS are ‘unregulated’ overstates the case: the principles of distinction, proportionality, precaution, and the Martens Clause apply to every weapon a State may lawfully develop or deploy, autonomous or otherwise, and the International Court of Justice’s Nuclear Weapons jurisprudence confirms IHL’s capacity to absorb novel military technology within its existing conceptual vocabulary. But the application of that vocabulary to systems capable of independent target selection exposes a genuine adequacy deficit in the institutional and evidentiary machinery through which IHL’s substantive standards are verified and enforced. Sovereignty concerns the reluctance of major military powers to accept external verification of weapons that may confer a decisive strategic advantage have thus far outpaced the humanitarian imperative identified by the UN Secretary-General and the ICRC.
FREQUENTLY ASKED QUESTIONS
Q1. Are Autonomous Weapon Systems illegal under current international law?
No. No treaty categorically prohibits AWS as a class of weapon. States retain the burden of ensuring that any AWS they develop or deploy can be used in conformity with IHL, principally through the Article 36 weapons-review obligation.
Q2. What is the ‘accountability gap’ and why does it matter?
It refers to the structural difficulty of attributing individual criminal responsibility or clear State responsibility when an autonomous system, rather than a human, causes an unlawful engagement.
