Lethal Autonomous Weapons Systems and the Governance Deficit: Challenges to International Humanitarian Law and Accountability
The rapid advancement of 'Lethal Autonomous Weapons Systems' (LAWS) has exposed significant gaps in the governance of emerging military technologies. This article examines how autonomous targeting challenges the principles of international humanitarian law, creates an accountability gap by reducing meaningful human control over lethal decisions, and outpaces existing regulatory frameworks. It argues that the current reliance on soft-law mechanisms within the 'Convention on Certain Conventional Weapons' (CCW) is inadequate, underscoring the need for stronger international governance to ensure legal accountability and preserve the integrity of the laws of armed conflict.
Introduction
The emergence of ‘Lethal Autonomous Weapons Systems’ (LAWS) platforms capable of selecting and engaging targets without meaningful human control represents one of the most consequential and least governable developments in the contemporary conduct of war. Unlike previous transformative military technologies, whose proliferation occurred within at least partial frameworks of international humanitarian law (IHL), LAWS challenge the foundational architecture of the laws of armed conflict at the point of their deepest assumption: that a human being, capable of moral and legal judgment, bears ultimate responsibility for the decision to kill. As the technological frontier advances- driven by competitive investment in artificial intelligence, machine learning, and autonomous systems by the United States, China, Russia, Israel, South Korea, and a growing number of second-tier military powers- the international community's governance response has remained stubbornly confined to the register of non-binding dialogue.
Discussions under the Convention on Certain Conventional Weapons (CCW) have produced a decade of expert deliberation without a legally binding outcome. A Group of Governmental Experts (GGE) has met repeatedly since 2014, producing guiding principles and recommending further study while declining to recommend a prohibition or regulatory treaty. Meanwhile, systems that meet or approach the functional definition of LAWS have been deployed in active conflict environments- including in Libya, Nagorno-Karabakh, and Ukraine- generating operational precedents that are outpacing normative development at an accelerating rate.1
This article examines the governance deficit surrounding LAWS through three analytical lenses: the IHL compliance challenges posed by autonomous targeting; the accountability gap created by the removal of meaningful human control from lethal decisions; and the structural obstacles- rooted in the strategic interests of leading military powers- that have prevented the emergence of an effective binding legal framework. The article argues that the 'soft law' approach pursued within the CCW process, while not without value, is structurally insufficient to address the risks that LAWS pose to the prohibition on arbitrary killing, the principle of distinction, and the long-run stability of the laws of armed conflict as a normative system.
Defining the Problem: What Are Lethal Autonomous Weapons Systems?
Conceptual and Technical Boundaries
The absence of an agreed definition of LAWS is itself a significant governance problem. Different actors have advanced definitions that serve their strategic interests: states with advanced autonomous weapons programmes have favoured narrow definitions that exclude their existing or near-term systems, while humanitarian organisations and smaller states have advocated broader formulations that centre the concept of 'meaningful human control' over targeting decisions.2
The International Committee of the Red Cross (ICRC) has proposed that LAWS be defined as autonomous weapons systems that can select targets and apply force against them without human intervention. This functional definition encompasses a spectrum from fully autonomous systems- where the human role is confined to system activation prior to deployment- to what scholars term 'human-on-the-loop' configurations, in which a human operator can observe and, in principle, override autonomous targeting decisions but is not required to authorise each individual engagement.3 The distinction matters because many systems currently deployed or under development fall into this intermediate category, and their legal status under existing IHL depends substantially on how 'meaningful human control' is operationalised.
Technologically, the capabilities that make LAWS possible- computer vision, sensor fusion, machine learning-based target recognition, and real-time decision algorithms- are not discrete military inventions but applications of broadly diffused civilian technologies. This dual-use character complicates governance: restrictions on LAWS development cannot be cleanly separated from restrictions on AI research more broadly, creating economic and scientific interests that resist regulatory demarcation. The asymmetry between the speed of technological development and the pace of normative response is therefore not merely a contingent failure of diplomatic will but a structural feature of the governance challenge.4
The Proliferation Landscape
The proliferation trajectory of LAWS-relevant technologies is already well advanced. The United States' Third Offset Strategy, launched in 2014 and its successors, explicitly identified autonomous systems as a central pillar of future military superiority, driving substantial investment in programmes including the Collaborative Combat Aircraft initiative and autonomous undersea vehicles.5 China's military-civil fusion strategy has similarly prioritised AI-enabled autonomous systems across air, maritime, and ground domains, with the People's Liberation Army identifying 'intelligentisation'- the integration of AI into all aspects of military operations- as a defining objective of modernisation through 2035.
Beneath the major-power competition, a second tier of proliferation is occurring among states and non-state actors with access to commercially available drone technology and open-source autonomy software. The Kargu-2 loitering munition, deployed by Turkish-backed forces in Libya's civil conflict, was identified in a 2021 UN Panel of Experts report as potentially having conducted autonomous attacks on logistics convoys- a finding that, if accurate, would represent the first confirmed use of a fully autonomous lethal system in combat.6 The barriers to entry for autonomous lethal capability are declining rapidly, and the assumption that LAWS governance primarily concerns great-power military programmes systematically underestimates the proliferation risk among technologically sophisticated non-state actors.
International Humanitarian Law and the Compliance Gap
The Principles of Distinction and Proportionality
IHL's application to LAWS is structured around two central principles: distinction and proportionality. The principle of distinction requires that parties to an armed conflict at all times distinguish between combatants and civilians, and between military objectives and civilian objects, directing attacks only against legitimate military targets. The principle of proportionality prohibits attacks expected to cause civilian harm that is excessive in relation to the concrete and direct military advantage anticipated.7
Both principles impose demands on the targeting decision that existing LAWS capability cannot demonstrably satisfy. Distinction, in the context of contemporary asymmetric conflict, requires contextual judgment of a kind that resists reduction to sensor data and pattern recognition algorithms. The distinction between a combatant carrying a weapon and a civilian farmer carrying an agricultural implement; between a military commander and a civilian official in a dual-use facility; between an active threat and a surrendering soldier- these judgments require the integration of contextual, cultural, and situational knowledge that current machine learning systems, trained on historical data, cannot reliably generalise to novel operational environments.8
Proportionality poses an analogous challenge. The assessment of anticipated civilian harm requires predictive judgment about the likely effects of an attack in a specific operational context, an assessment that involves not merely computational optimisation but normative weighing of incommensurable values. The ICRC has consistently maintained that proportionality assessment requires a 'human value judgment' that cannot be delegated to a machine, a position supported by a substantial body of legal scholarship and the stated positions of a majority of states in the CCW GGE process.9
The Precautionary Principle and Target Verification
Beyond distinction and proportionality, IHL's precautionary principle requires that parties take all feasible precautions in the choice of means and methods of attack to avoid or minimise civilian casualties. In the LAWS context, this requirement intersects with the technical question of target verification: the degree of certainty with which an autonomous system can confirm the military nature of a target before engagement. Current computer vision systems, even those trained on large datasets, exhibit failure modes including misidentification under adverse conditions, adversarial manipulation- where targets are deliberately altered to deceive recognition systems- and systematic bias derived from non-representative training data.10
The operational environments in which LAWS are most likely to be deployed- complex urban terrain, asymmetric conflict with irregular combatants, information-degraded environments- are precisely those in which the failure modes of autonomous target recognition are most acute. The precautionary principle's requirement for feasible verification thus points toward a structural tension between the operational utility of LAWS and their IHL compliance, a tension that technology optimists who argue that future AI capabilities will eventually satisfy IHL requirements have not yet resolved with empirical evidence.
The Accountability Gap
The 'Responsibility Gap' in Legal Theory
The accountability dimension of the LAWS governance problem has been most systematically developed through the concept of the 'responsibility gap,' introduced into the academic literature by Robert Sparrow and subsequently elaborated by legal scholars including Christof Heyns and Human Rights Watch's Bonnie Docherty.11 The responsibility gap identifies a structural problem: when an autonomous system makes a targeting decision that results in unlawful killing, the existing architecture of individual criminal responsibility under international law- which requires proof of intent, knowledge, and causal agency- may leave no individual legally accountable for the act.
The potential nodes of accountability, the programmer who designed the algorithm, the commanding officer who authorised deployment, the political official who approved the system's acquisition- each face evidentiary and legal obstacles to prosecution that autonomous decision-making systematically amplifies. The programmer did not direct the specific targeting decision. The commanding officer may not have been able to foresee the specific context of engagement. The political official exercised no operational control. The result is not merely impunity in individual cases but a structural erosion of the deterrent function of international criminal law in the context of autonomous warfare.12
State Responsibility and Command Responsibility
At the level of state responsibility, the deployment of LAWS creates challenges for the application of Articles 4 and 8 of the International Law Commission's Articles on State Responsibility, which attribute conduct to states on the basis of direction and control. An autonomous system that acts in ways not specifically directed by state agents- including in ways that exceed the parameters of its deployment authorisation creates novel questions about the causal chain between state decision and harmful outcome.13
Command responsibility- the doctrine that military commanders bear criminal liability for war crimes committed by forces under their effective control, if they knew or should have known of the crimes and failed to prevent or punish them- similarly strains when applied to autonomous systems. The effective control standard presupposes a capacity for continuous oversight and intervention that 'human-on-the-loop' LAWS configurations do not guarantee. A commander who deploys an autonomous system with a known rate of target misidentification and accepts the resulting civilian casualties as an anticipated operational parameter occupies a morally and legally ambiguous position that existing doctrine was not designed to address.14
The Governance Response and Its Structural Limits
The CCW Process and the Failure of Consensus
The primary multilateral forum for LAWS governance has been the CCW GGE, which has operated since 2014 under a consensus rule that effectively gives any major military power a veto over binding outcomes. The eleven guiding principles adopted by the GGE in 2019- affirming the applicability of IHL, the importance of human responsibility, and the need for accountability- represent the ceiling of consensus achievable within the forum, and fall well short of the prohibition or legally binding regulation that a majority of participating states have endorsed.15
The structural obstacle to binding regulation within the CCW is transparent: the United States, Russia, and China- the three states with the most advanced autonomous weapons programmes- have each, in different ways, resisted proposals for binding constraints. The United States has consistently argued that existing IHL is sufficient to govern LAWS and that premature regulation risks constraining technological development that offers military and, it argues, humanitarian advantages. Russia has questioned whether LAWS can be meaningfully defined. China has expressed support for a political declaration against fully autonomous weapons while resisting legally binding constraints on development and testing.16
The result is a process that produces procedural continuity without substantive progress- a form of governance that legitimises the status quo of unconstrained development by providing the appearance of international deliberation. This dynamic, in which a multilateral forum is captured by the interests of the states most invested in the technology under discussion, is not unique to LAWS but is particularly consequential given the pace of autonomous weapons development relative to the timeline of diplomatic consensus-building.
Civil Society, the ICRC, and the Push for a Binding Treaty
Outside the CCW, a significant coalition of states- currently numbering over seventy, including Austria, Chile, Costa Rica, Mexico, New Zealand, and the majority of African Union members- has advocated for a legally binding instrument to prohibit fully autonomous weapons and require meaningful human control over targeting decisions in all circumstances.17 This coalition has been supported by the Campaign to Stop Killer Robots, a civil society coalition of over two hundred organisations, and by successive ICRC statements calling for new binding rules.
The humanitarian law case for a prohibition treaty draws explicitly on the precedent of the Ottawa Treaty on anti-personnel landmines (1997) and the Convention on Cluster Munitions (2008), both of which were achieved through processes that bypassed the consensus constraints of the Conference on Disarmament by convening willing states outside established multilateral forums. Whether the political conditions for a similar 'Oslo Process' approach to LAWS exist is uncertain: the campaign against landmines and cluster munitions benefited from a relatively clear humanitarian harm profile and the absence of great-power strategic interest in the prohibited weapons. LAWS, by contrast, are central to the military strategies of the leading powers, making exclusion of those powers from a treaty process a more significant limitation on its effectiveness.18
Toward a More Effective Governance Framework
The inadequacy of the current governance response does not foreclose the possibility of more effective frameworks, though it does constrain the range of politically achievable options. Three directions merit consideration.
First, a shift in the locus of norm development from the CCW to the UN General Assembly- where consensus rules do not apply and resolutions can be adopted by majority- offers a pathway to establishing the political and normative groundwork for a treaty process. A General Assembly resolution affirming that fully autonomous weapons are incompatible with existing IHL, while not legally binding, would represent a significant shift in the international normative environment and could isolate the major powers' resistance more visibly within global public discourse.19
Second, the meaningful human control standard- while contested in its definition- offers a technically operationalisable threshold that could form the basis of a binding regulatory instrument short of full prohibition. Requiring that human operators make affirmative engagement decisions for each target, with sufficient time and information to exercise genuine judgment, would constrain the most dangerous applications of autonomous targeting while permitting autonomous functions in non-lethal domains. The challenge of verification and enforcement would be substantial but not categorically different from verification challenges in other arms control contexts.20
Third, domestic legal and procurement frameworks represent an underutilised lever of governance. States that have articulated domestic policies on autonomous weapons- including the United States' DoD Directive 3000.09- have demonstrated that operational standards can be codified in regulatory form. Harmonising such domestic frameworks, and creating transparency mechanisms that allow assessment of compliance, could generate a de facto international standard with meaningful practical effect even in the absence of a formal treaty.21
Conclusion
Lethal autonomous weapons systems present a governance challenge that is simultaneously legal, ethical, and strategic. The existing international humanitarian law framework was not designed for targeting decisions made without human judgment, and the accountability structures of international criminal law and state responsibility cannot straightforwardly accommodate the responsibility gap that autonomous lethal action creates. The CCW process has provided a forum for deliberation, but not a mechanism for binding constraint, and the pace of technological development among leading military powers is generating operational facts that are progressively narrowing the space for preventive regulation.
The case for a binding international instrument- whether a prohibition on fully autonomous weapons or a mandatory meaningful human control standard- rests not only on the humanitarian risks of unconstrained LAWS deployment but on the systemic risk to the laws of armed conflict as a normative order. A regime of IHL whose foundational principles can be systematically evaded through the interposition of autonomous decision-making between human agency and lethal outcome is a regime whose long-run legitimacy and effectiveness are in question. The governance response to LAWS is therefore not merely a discrete arms control problem but a test of whether the international community retains the collective capacity to subordinate technological possibility to legal and ethical constraint.
References:
1. Asaro, P. (2012). On banning autonomous weapon systems: Human rights, automation, and the dehumanization of lethal decision-making. International Review of the Red Cross, 94(886), 687–709. https://doi.org/10.1017/S1816383112000768
2. Bode, I., & Huelss, H. (2018). Autonomous weapons systems and changing norms in international relations. Review of International Studies, 44(3), 393–413. https://doi.org/10.1017/S0260210517000414
3. Campaign to Stop Killer Robots. (2023). Country views on killer robots. https://www.stopkillerrobots.org/wp-content/uploads/2023/09/KRC_CountryViews2023.pdf
4. Docherty, B. (2012). Losing humanity: The case against killer robots. Human Rights Watch. https://www.hrw.org/report/2012/11/19/losing-humanity/case-against-killer-robots
5. Docherty, B. (2020). Mind the gap: The lack of accountability for killer robots. Human Rights Watch. https://www.hrw.org/report/2015/04/09/mind-gap/lack-accountability-killer-robots
6. Ekelhof, M. (2019). Lifting the fog of targeting: 'Autonomous weapons' and human control through the lens of targeting. Naval War College Review, 71(3), 61–96. https://digital-commons.usnwc.edu/nwc-review/vol71/iss3/6/
7. Etzioni, A., & Etzioni, O. (2017). Pros and cons of autonomous weapons systems. Military Review, 97(3), 72–81. https://www.armyupress.army.mil/Portals/7/military-review/Archives/English/MilitaryReview_20170630_art009.pdf
8. Heyns, C. (2013). Report of the Special Rapporteur on extrajudicial, summary or arbitrary executions. United Nations Human Rights Council, A/HRC/23/47. https://www.ohchr.org/Documents/HRBodies/HRCouncil/RegularSession/Session23/A-HRC-23-47_en.pdf
9.Human Rights Watch & International Human Rights Clinic. (2016). Making the case: The dangers of killer robots and the need for a preemptive ban. Human Rights Watch. https://www.hrw.org/report/2016/12/09/making-case/dangers-killer-robots-and-need-preemptive-ban
10. International Committee of the Red Cross. (2021). ICRC position on autonomous weapon systems. https://www.icrc.org/en/document/icrc-position-autonomous-weapon-systems
11. International Committee of the Red Cross. (2023). Autonomous weapons: Hard limits are necessary and achievable. https://www.icrc.org/en/document/autonomous-weapons-hard-limits-necessary-achievable
12. Johnson, J. (2021). Artificial intelligence & future warfare: Implications for international security. Defense & Security Analysis, 35(2), 147–169. https://doi.org/10.1080/14751798.2019.1600800
13. Meier, O. (2021). Autonomous weapons and the limits of analogy: Lessons from the landmine and cluster munitions treaties. Nonproliferation Review, 28(1–3), 43–60. https://doi.org/10.1080/10736700.2021.1984444
14. Roff, H. M. (2014). The strategic robot problem: Lethal autonomous weapons in war. Journal of Military Ethics, 13(3), 211–227. https://doi.org/10.1080/15027570.2014.975010
15. Schmitt, M. N. (2013). Autonomous weapon systems and international humanitarian law: A reply to the critics. Harvard National Security Journal Features. https://harvardnsj.org/wp-content/uploads/2013/02/Schmitt-Autonomous-Weapon-Systems-and-IHL-Feature.pdf
16. Schmitt, M. N., & Thurnher, J. S. (2013). 'Out of the loop': Autonomous weapon systems and the law of armed conflict. Harvard National Security Journal, 4(2), 231–281. https://harvardnsj.org/wp-content/uploads/2013/01/Schmitt-and-Thurnher-Final.pdf
17. Sharkey, N. (2010). Saying 'no!' to lethal autonomous targeting. Journal of Military Ethics, 9(4), 369–383. https://doi.org/10.1080/15027570.2010.537903
18. Sparrow, R. (2007). Killer robots. Journal of Applied Philosophy, 24(1), 62–77. https://doi.org/10.1111/j.1468-5930.2007.00346.x
19. United Nations Group of Governmental Experts on Lethal Autonomous Weapons Systems. (2019). Guiding principles affirmed by the Group of Governmental Experts on emerging technologies in the area of lethal autonomous weapons systems. CCW/GGE.2/2019/CRP.1. https://www.unog.ch/80256EDD006B8954/(httpAssets)/BE13D84C23A7B3B4C12584E0003E4F9E/$file/CCW_GGE.2_2019_CRP.1_Rev.2.pdf
20. United Nations Panel of Experts on Libya. (2021). Final report of the Panel of Experts on Libya established pursuant to resolution 1973 (2011). S/2021/229. https://www.undocs.org/S/2021/229
21. U.S. Department of Defense. (2023). DoD directive 3000.09: Autonomy in weapon systems. https://www.esd.whs.mil/Portals/54/Documents/DD/issuances/dodd/300009p.pdf
22. Vestner, T., & Rivera, J. (2019). Autonomous weapon systems and IHL: The question of meaningful human control. Global Policy, 10(3), 329–337. https://doi.org/10.1111/1758-5899.12668
23. Wareham, M. (2020). Banning killer robots: The 2019 Human Rights Watch world report. Human Rights Watch. https://www.hrw.org/world-report/2020/country-chapters/arms
24. Zwitter, A., & Gstrein, O. J. (2020). Big data, privacy and COVID-19- Learning from humanitarian expertise in data protection. Journal of International Humanitarian Action, 5(1), 4. https://doi.org/10.1186/s41018-020-00072-6
(The views expressed are those of the author and do not represent the views of CESCUBE)
Photo by Simon Fitall on Unsplash
