Drone Warfare and the Destruction of Gaza
Part 2 – Drone Swarms and the Unmaking of International Humanitarian Law
Shiri Krebs
‘Prove you’re human.
Prove you stand on two legs.
Run.’
(from ‘Running Orders’, by Lena Khalaf Tuffaha)
From Ukraine to Iran, to Lebanon and to Palestine, drone technologies and digital systems are used to generate, prioritize, and select targets for military actions, inflicting violence on people and places. These systems have a central role in both escalating and legitimizing military violence. The first part of this essay (available here) examined how drone sensing and swarms’ capabilities erase Palestinians’ personhood. In this second part of my essay, I explain how these digital capabilities change the application and meaning of core International Humanitarian Law (IHL) norms in Gaza and beyond, unmaking established limitations on organized armed violence. I focus on three issues that are central to drone warfare and demonstrate how drone technologies weaken the protective power of IHL. First, drone technologies change practices of legal interpretation and meaning-making, by shifting standards and adopting codifiable definitions. Second, drone technologies change the scale of violence, producing information that can justify numerous concurrent attacks while masking the implications of the overall campaign. Third, drone technologies generate an accountability gap, as machines lack intention and decisions are fragmented into many opaque choices and actions.
Drone technologies change practices of legal interpretation and meaning-making
Legal terms in the digital age need to be codifiable or explained in terms that can be processed by digital systems. For example, a legal term such as ‘high-level target’ must be codified and translated to something calculable, and the term ‘civilian’ must be translated to elements that can be processed by algorithms and a technical lens. Even simpler concepts, like ‘child’, which are relevant to the legal classification and scope of protection afforded to people during armed conflicts, must be codifiable and require programming determinations concerning physical attributes and other relevant factors. The codification process reduces these complex and context-dependent human concepts to something different, technical in nature, less sensitive to social contexts and varying situations.
Additionally, core legal standards are designed for human ethics and cognition and are incompatible with decisions made by machines. For example, common legal standards include requirements that a decision is made ‘in good faith’, with an ‘acceptable doubt’, or based on ‘reasonable belief.’ But machines do not exercise ‘doubt’ and do not have beliefs, reasonable or not. The concept of ‘good faith’ is also foreign to their programming and processing. This incompatibility means that core legal standards concerning evidence and proof cannot be complied with, and are, instead, replaced by codifiable alternatives and technical notions of error rates and mechanical calculations.
Drone technologies change legal sensibilities
Drone technologies change the scale of military violence based on constant surveillance capabilities and probabilistic inferences about people and places. This changes the context in which legal norms function, further detaching IHL from its inherent aim of protecting people from the harms of war. For example, targeting algorithms can rapidly generate numerous targets, significantly increasing the number of possible operations and attacks. The capabilities of drone swarms and AI-powered targeting tools, along with the types of data they collect on people and places (‘potential targets’), drive a shift from few to many targets and from high- to low-level targets. In turn, legal terms, such as ‘high-level target’, are also recalibrated through codification processes as noted above. These changes legitimize more attacks and, by definition, tolerate more collateral damage as acceptable, simply because of the rise in the number and scale of legitimized operations. AI-powered targeting algorithms further legitimize erroneous deaths as an inherent part of the system, based on their declared error rates. If a targeting algorithm’s known error rate is 10%, this means that by definition, out of a hundred people (or places) targeted, we know in advance that 10 are innocent and will be wrongly killed, accepting these unjustified deaths ex ante as an inherent part of the system. Paradoxically, the arbitrariness of these systems, their lack of intent, and the distance they create between military violence and human decision-makers make it easier to legitimize attacks and to characterize harm inflicted on innocent civilians as a technical error or an unintended consequence of the system. Together with the fact that drone swarms significantly reduce the risk to those operating them and enable the generation of more targets, this practice drives an escalation of violence and destruction.
This escalation in destruction is also known as the precision paradox. A precision paradox occurs when the use of more precise or sophisticated weapons systems, such as drone swarms, doesn’t actually reduce harm and may even intensify destruction. In the case of drone swarms, an increase in harm may occur – even if the targeting is accurate and the inherent error rate is low – simply by the improved capabilities to generate many targets at speed, legitimizing violence at a greater scale. The use of constant, overarching surveillance for threat detection continuously generates humans as threats and presents their elimination as both urgent and necessary.
The AI-powered targeting systems currently applied by the Israeli military in Gaza exemplify this dynamic. While the military acknowledged the use of some AI systems, such as ‘the Gospel’ (‘Habsora’, in Hebrew), in its targeting practices, other systems, such as ‘Lavender’, were exposed by whistle-blowers. The reports (including information provided by the military) clearly show a dramatic increase in the number of AI-generated targets, and minimal human intervention or review (as a direct result of the massive increase in target generation). The outcomes of these targeting systems are witnessed in the devastation in Gaza, and in the dramatically high percentage of women and children among the victims. In the AI-generated assembly line of human targets in Gaza, Palestinians are ‘framed’ as targets and killed with little human involvement. Most deaths are never investigated or accounted for, and meaningful information about the targeting process and the reasons for and circumstances of the killing is often unavailable, preventing effective legal evaluation.
Of course, Gaza is not the only context in which drone warfare results in intense violence and harm to civilians. Both Russia and Ukraine have been deploying AI-powered drones. These systems, according to The Lancet, caused ‘more casualties in Ukraine than any other weapon.’ In Iran, an unknown system (probably Maven), identified Shajareh Tayyebeh primary girls’ school in Minab as a military target, likely based on outdated information and proximity to military barracks. The attack killed about 175 people, most of them young schoolgirls, between the ages of seven and twelve.
While it is unclear whether AI was involved in the decision-making process that led to the horrific attack on the Minab Primary School (and if so, to what extent), it is clear that a complex assemblage of digital systems is compressing and obscuring the targeting process, generating more targets at a quicker pace than ever, based on opaque technical processes. This compression speeds up the targeting process significantly, limiting the role of humans in the targeting process. As humans, we cannot process information or make informed judgments at the speed that our war machines generate and select new targets. The digitized target-generation process is not only too fast for human cognition and too voluminous and complex for human comprehension; it changes how we view ‘people’ and ‘targets,’ what targets are detected and prioritized, and how the gravity of threats – and the legality of planned operations – is assessed.
In Ukraine, Iran, and Gaza, drone technologies are not merely changing the conduct of warfare; they are reshaping legal sensibilities, normalizing intensified violence through techno-legal frameworks that become increasingly data-driven, procedural, and detached from lived human experience.
Drone technologies lack intention, reflection, and remorse
Drone technologies create a profound challenge for legal concepts of intent, responsibility, and accountability. Legal accountability – including individual criminal responsibility – rests on the assumption that harmful acts can be traced to human decision-makers whose intentions can be assessed and judged. Yet machines cannot form intentions, exercise moral judgment, or bear responsibility. And the centrality of drone technologies in military decision-making processes distances the humans-in-the-loop from the decisions and their outcomes.
This creates particular difficulties for crimes such as genocide, where liability depends upon proof of a specific intent to destroy, in whole or in part, a protected group. While genocide scholars have long recognized the difficulties of inferring collective or state intent, drone technologies further complicate this task by fragmenting decisions across complex socio-technical systems involving operators, commanders, software engineers, algorithms, sensors, and autonomous processes.
The result is not simply an evidentiary challenge, but a transformation in how violence is experienced and understood. From the perspective of those subjected to drone attacks, destruction often appears arbitrary, opaque, and devoid of human agency. Deaths frequently go uninvestigated, meaningful information about targeting decisions remains inaccessible, and avenues for redress are limited or nonexistent. As decisions become increasingly mediated by screens, data streams, and automated systems, responsibility becomes more diffuse and difficult to locate. The drone lens thus widens the accountability gap in contemporary warfare, distancing human actors from the consequences of military actions while leaving victims exposed to harm that appears to come from nowhere and from everywhere, and that is answerable to no one.
Shiri Krebs is a Professor of Law and the Director of the Centre for Law as Protection at Deakin University, Melbourne, Australia. She is also an Experienced Alexander von Humboldt Research Fellow at Hamburg University, and an Affiliate Scholar at the Stanford Center for International Security and Cooperation (CISAC). This essay is based on research funded by the Australian government through the Australian Research Council (DE230101646).