Alongside the urbanization of armed conflict lies a second trend: the increase in the use of technology to augment methods of warfare. Although some may claim that the integration of technological solutions – such as increasing the precision of munitions or reliance upon enhanced data and surveillance collection and analysis tools – can reduce civilian harm, contemporary urban conflict demonstrates that more should be done to ensure that new weapons technologies incorporate additional measures to minimize civilian harm when deployed in urban environments.

In this post, part of a special series on urban warfare, Damian Copeland and Lauren Sanders, both senior research fellows at the University of Queensland Law and Future of War Project, propose that enhancing existing IHL obligations in reviewing weapons, means and methods of warfare, by adopting a cooperative approach between industry and States, promises to help bridge this gap. The result is both business efficiency in the design and development of new technology, but more importantly – harmonization and increased compliance with IHL for new technologies likely to be fielded during urban warfare.

ICRC Humanitarian Law & Policy Blog · Engaging with the industry: integrating IHL into new technologies in urban warfare

During his opening remarks to the 33^rd International Conference, Peter Maurer noted with concern that urban warfare has become the ‘new normal’. The human cost of urban warfare, measured in death, injury, permanent disability, displacement, reduced access to services, and psychological trauma, is high. The world has repeatedly witnessed with horror the devastation and suffering caused during intense urban battles, which – despite the existence of international humanitarian law (IHL) aimed at protecting the civilian population – has not abated in contemporary conflicts. In fact, the impact of urban warfare is amplified by the trend of increasingly urbanized populations and the renewed strategic significance of populated areas. Further, there has been an increase in non-State actors’ participation in conflict, who, because of resourcing issues and the nature of their organization, tend to utilize civilian infrastructure by default for their military purposes.

Developments in warfare technologies have further modified the impact of urban warfare upon civilian populations. For example, sustained aerial strikes using precision munitions were critical to ground operations against Islamic State group forces in Iraq, and in particular during operations in Mosul, in 2017, but still resulted in catastrophic damage to the civilian population. In contrast, during the 2020 Nagorno-Karabakh conflict in Azerbaijan, the heavy reliance on the use of drones to support strike operations has been lauded as the ‘future of war’, based upon the suggestion that the use of drone surveillance and weapon systems enabled a swifter and more targeted conclusion of fighting by more readily identifying and then destroying military objects; without it, the argument goes, the fighting would have become protracted had those military capabilities had time to embed themselves in dense urban areas of the city.

There are two competing views on the role of technology, such as drones (or uncrewed aerial vehicles) and precision guided munitions, can have in terms of impact upon civilians during urban warfare. The first is that the import of precision technologies and enhanced intelligence surveillance and reconnaissance systems can reduce harm to civilians by enabling better distinction, more accurate collateral damage estimation and the use of lower-yield explosive munitions to achieve the same military effect. Using multiple sensors to provided additional information to support decision making; the correlation of large swathes of data using algorithms to quickly sort, categorize and interpret operational circumstances; and the operation of guidance technologies to enhance predictability of weapon system accuracy, offer the potential to reduce civilian harm during urban conflict. Such technologies can support the good faith application of IHL rules relating to distinction, proportionality and precautions, including through the provision of more complete intelligence and real time surveillance feeds. In turn, this can reduce the risk of targeting errors due to a reliance on inaccurate or incomplete information.

The contrary view suggests that such technological developments expose civilians to greater harm, as it encourages the use of weapons in urban areas where the effects of less precise munitions would have otherwise been considered unlawful. Specifically, the ‘precision paradox’ enables the use of lower-calibre precision munitions to change the proportionality calculus for individual attacks. This invites more frequent use of force in urban areas against military objectives where less precise munitions would have failed to achieve a proportionate use of force having regard to the assessed military advantage, or would have violated the prohibition against indiscriminate attacks. On the latter view, the cumulative effect of precision attacks over the course of an urban battle can generate a ‘rolling wave of destruction’ that can devastate a city.

Regardless of which view you accept, the incorporation of new technologies to enhance military force capabilities and their use in urban warfare appears to be an increasing trend. Equally, their adoption by non-State actors is more prevalent where such technology is seemingly ubiquitous. Consequently, the engagement of international humanitarian law considerations during their design, development and acquisition phases, through governmental cooperation with defence industries, is crucial to ensuring that when such technologies are adopted, they, by design and default, ‘build in’ humanitarian law considerations that mitigate, to the extent possible, the impact upon civilian populations use of these technologies in increasingly urbanized conflict.

This obligation to determine compliance with IHL in the study and development phases of new weapons, means and methods of warfare is already applied by States, through the conduct of Additional Protocol I Article 36 weapons reviews, but compliance is not uniform in practice. Moreover, the weapon review process determines a minimum standard of legal compliance and does not seek to fully exploit the potential benefits of technology in reducing harm to civilians during urban warfare. Weapon reviews enable States to not only determine the legality of a weapon but also to identify how technology can enable potential humanitarian benefits through industry and defence cooperation in the development of AI systems. Accordingly, enhancing compliance with this process, and focusing upon iterative design engagement with industry during design of these technologies can and should enhance harm minimization when these technologies are eventually adopted by States for use in armed conflict.

AI technology that could enhance IHL compliance in urban warfare

Technology may be part of a broader solution that enables States and non-State actors to strengthen the protection of civilians by reducing civilian harm, including harm caused by high explosive weapons in urban environments. AI can provide belligerents with more accurate, reliable information that can better inform distinction and proportionality decisions, what precautions might be feasible in the circumstances, and potentially reduce the use of explosive weapons with wide area effects. When better information is available to inform targeting decisions, the result should be better protection of the civilian population. Consider for example, these two seemingly obvious situations where AI can be used for reducing civilian harm: (1) the use of AI image recognition technology, similar to that used in medical and automotive application, to recognize and track the location of the civilian population in urban settings; and (2) the use of image recognition to identify protected emblems including the red cross, red crescent, civil defence and cultural property.

1. Recognizing and tracking the location of the civilian population

A tragic example of the use of wide area explosive weapons in urban warfare occurred at 8am on 17 March 2017, during the US-led campaign to help Iraqi forces retake the city of Mosul from Islamic State Group (ISG) fighters. The US conducted an airstrike on two ISG fighters reportedly firing at Iraqi soldiers from the roof of a civilian building in Al Jadida, which was housing up to 150 displaced Iraqi civilians inside. It was estimated that only six civilians survived. It is unclear if those civilians were in that building voluntarily seeking shelter, or because of a tactic of Islamic State Group to crowd civilians into buildings being used for military purposes, and whether or not they were given a warning to evacuate the building. However, their presence did not appear to be known to either the US or Iraqi forces prior to the strike, but was known to the Islamic State Group.

While the location of civilians was not observed by coalition drone footage on this occasion, such footage is often recorded, enabling civilian movements on the battlefield to be monitored. These systems are typically monitored by human operators manually plotting civilian movements, based upon limited available non-persistent intelligence, surveillance and reconnaissance assets, that are also tasked with attempting to monitor movements of opposition forces. It is conceivable that had a more persistent civilian tracking technology been in place, there may have been a better general understanding of the location of civilians in and around the area surrounding this fighting location.

We are not suggesting that the use of such systems is a technical panacea that will eliminate civilian casualties from urban warfare; rather that integrating more advanced technologies specifically designed with the idea of reducing civilian harm is crucial to reducing civilian harm during urban conflict. Existing technological and physical limits when relying on drone footage can result in over-reliance upon technology to guide decisions to use force. In this case, the limited capacity to observe civilian population movements around Al Jadida due to poor weather conditions in the two days preceding the strike may be overcome by future technology developments in this field. Equally, data bias and the inherent limitations on emerging technologies’ ability to fully address the complexities of an urban environment mean that the value of these technologies is best embraced when properly integrated into existing decision-making processes, and with transparency about system limitations.

We contend that the use of more refined, and targeted technologies will assist commanders in making strike decisions in urban environments. Notably, such systems require careful design, testing and evaluation systems to ensure that they are fielded within limitations, and that those relying upon their outputs understand their limitations. For these reasons, integration with industry during the design and development phase, and the end-users of the technology – military decision makers – is critical.

The ability to programme and train AI to recognize patterns in large volumes of data has been recognized in a range of industries and applications including diagnosing illness in medical imaging. The same technology is also used to enhance enemy target recognition and can be applied to support armed actors in recognizing and recording the movement and location of civilians on the battlefield. An AI system designed to fuse video data feeds from a multitude of sources, including drones, and other aerial- and ground-based surveillance platforms can be used to inform military commanders who plan or decide an attack, enabling better distinction and proportionality decisions and ultimately reducing the risk of similar incidents.

Current challenges in surveillance relate to limited resources by relying upon small numbers of large and expensive drones, whereas the ‘small, smart and many’ approach may adjust the current system of ‘looking at the city through a straw’, to one that allows more persistent surveillance soaking, and thus a better understanding of the civilian population’s location and movements. Combined with software to track and digitally superimpose the data obtained from these systems on targeting maps or imagery, the risk of erroneous decisions in battle are reduced. Notably, this technology does not cure all problems of data bias, confirmation bias, cultural misunderstanding or human error; but provided the software and operators of the technology have appropriate reliability and accountability measures built into the design, testing, training and use parameters of this technology, it can serve as a meaningful check and balance to improve existing human decision-making processes.

2. Identifying distinctive emblems

IHL utilizes distinctive emblems to facilitate the visual identification of protected persons and objects such as medical personnel, transports and units, civil defence and cultural objects and places of worship. Unfortunately, this humanitarian objective is not always achieved. A well-known example occurred in 2015, when a US gunship attacked a Médecins Sans Frontières hospital in Kunduz, Afghanistan, killing 42 people and wounding 37. While the hospital’s GPS coordinates were known to the US and Afghan forces, the subsequent US investigation revealed that the air-crew did not consult their ‘no-strike list’ and experienced other technical failures, that, when coupled with human error, resulted in the misidentification of the hospital as a Taliban-held location. Although there were multiple errors – both human and technical – that cumulatively resulted in this tragedy, there was no technical circuit-breaker in use, which potentially could have served as an additional check and balance to aid in preventing such an attack.

Technology may enhance the lawful use of weapons by performing a governance role designed to ensure IHL compliance by protecting medical personnel and medical units from misidentification. For example, command and decision support tools, with the capability of integrating information from multiple sources, can be used to expand the information reasonably available to a commander to make decisions. Some argue a moral imperative to use minimally-just autonomy that performs tasks, such as preventing the misuse of weapons, to enhance ethical and legal standing, whereas the availability of such technology may even import a legal obligation to utilize it, if it is feasible to do so. If such technology existed, the layering of safeguards provided by using technology to counter contingency, imperfect information, and potentially reduce human tendencies of confirmation bias, could act to prevent the use of weapons in situations such as the events in Kunduz in the early morning of 3 October 2015.

Enhancement of existing review obligation: review-led design

A key requirement to foster innovation in technology that reduces the human cost of urban warfare is to engage early in the research and design of weapon-related technology. This includes State-sponsored research and development by private defence industry, research organizations and academic institutions. If an integrated approach to innovation was adopted by States, incorporating a State’s legal obligations earlier in the design and development phase of technology development, significant advances could be made to address the risks. This innovative approach has been acknowledged as one way to mitigate technologies of concern. For example, the Guiding Principles affirmed by the GGE on LAWS, specifically refer to the need to incorporate risk mitigation and IHL compliance measures in the design, development, testing and deployment cycle of emerging technologies in weapon systems. The result may contribute to better human decisions in the use of force across the entire research, development and use of the weapon system.

We argue that the existing legal obligation of Article 36 review of new weapons – particularly during the early design and development phase – provides an opportunity for States to study and develop technology that not only complies with the reviewing State’s IHL obligations but also enhances IHL compliance by design to reduce the risks to civilians during urban warfare. While the focus of the traditional Article 36 review occurs at the conclusion of its design and development phase, a review-led design approach would shift focus to the design and development phase to inform innovation to actively consider how advances in computing power and availability of data can improve humanitarian outcomes.

What defence-industry engagement could look like

State practice suggests that a limited number of countries have enacted a formal Article 36 review mechanism. Moreover, State practice in terms of completion of weapons reviews and their content is often shielded from public view for security reasons. One public example, however, is the Australian position on the conduct of weapons reviews and the enhancement of sovereign capabilities that incorporate legal and ethical design considerations, espoused in a recently published Defence Science Technology report, A Method for Ethical AI in Defence. In addition to the Australian Defence Force’s guide to weapons review processes, it establishes a process to engage Defence industry during the design and acquisition phase through an iterative discussion between the system designer and Defence, called the Legal and Ethical Assurance Program Plan (LEAPP).

The use of such a document encourages a discourse between the developing company and the user of the capability – who is ultimately responsible for the lawful use of the system. In addition to enhancing potential compliance with IHL, the LEAPP provides a level of business efficacy for engaged companies: they invest in the design requirements of their client, having regard to the final certification requirements for acquisition of the system. It further supports a State’s ability for early adoption of technology, and the current focus on industry cooperation to support militaries’ ‘capability edge’, as this built-in, not bolted-on approach supports certification of weapons systems for introduction into service without the addition of lengthy testing and evaluation processes, and may even avoid potential system redesign given possible compliance issues will be identified as they are being developed.

Cooperation with industry to give effect to the whole of development, design and use cycle incorporating IHL compliance is a win-win for militaries and industry, while enabling the promise of technology to deliver advancements in the more refined use of force in urban environments. Whether the promise of military AI has been oversold or not, any technology developed and designed with IHL in mind gives effect to ensuring capability is limited in use for legitimate purposes only, and through incorporating inherent design limitations that factor in IHL requirements, supports prevention of misuse of new or novel technology in order to achieve the aims of IHL. This supports what should be the ultimate aim of any fighting force – to end the conflict as swiftly as possible while minimizing harm to the civilian population.

Editor’s Note: Damian Copeland and Lauren Sanders are both Senior Research Fellows at the University of Queensland’s Law and the Future of War Project. This article represents the authors’ personal views and does not reflect those of the Australian Defence Force or Australian Government. The research for this paper received funding from the Australian Government through Trusted Autonomous Systems, a Defence Cooperative Research Centre.

AI technology that could enhance IHL compliance in urban warfare

1. Recognizing and tracking the location of the civilian population

2. Identifying distinctive emblems

Enhancement of existing review obligation: review-led design

What defence-industry engagement could look like

See also

Introduction