​​From Kill Switch to Peace Switch: Rethinking AI Governance in African Military Law

How to cite this journal: Author, Date of the post, WMO Conflict Insight, Title of the post, ISSN: 2628 6998, https://worldmediation.org/journal/

EDITORIAL NOTE *

This article names particular states in connection with surveillance practices and assesses the conduct of their governments. Those assessments reflect the author’s own perspective and rest on the sources cited; this journal’s editorial position is neutral. The article was written in 2024 and should be read with that date in mind.

ABSTRACT

The escalating integration of artificial intelligence into warfare has ignited a fierce global debate, often narrowly confined to the perils of killer robots, or lethal autonomous weapons systems, and to calls for outright bans. This framing, while highlighting critical ethical concerns regarding accountability and human control over decisions of life and death, risks obscuring the broader transformative potential of the technology and the need for more holistic governance. Humans remain superior to such systems in judgement and knowledge under significant uncertainty. This paper argues that focusing merely on a kill switch, meaning the prohibition or stringent control of autonomous weapons, is an insufficient and reactive approach to governance. It advocates instead a peace switch paradigm, urging a proactive re-evaluation of the diverse applications of this technology in order to foster international security and stability. Beyond autonomous targeting, it enhances intelligence, surveillance, reconnaissance, logistics, and decision support. Drawing on global efforts to regulate military applications, particularly the human-centred approaches developed in Europe, this paper argues that the dual-use nature of the technology, in which commercial advance significantly outpaces military development, requires integrated and forward-looking legal frameworks. While the sources consulted do not specifically address African military law, the principles explored here offer a foundation for any region seeking to move beyond a purely prohibitive stance. Such an approach would embed human-centred ethical guidance within comprehensive legal and policy frameworks, ensuring that the evolution of this technology serves not only to avert catastrophic conflict but actively builds pathways to lasting peace and security.

KEYWORDS

Artificial intelligence, humanitarian, drones, peace, law, military

INTRODUCTION

The Harmattan, a relentless wind whipping sand across the vast plains of the Sahel, carries with it a new kind of chill. It is the chill of uncertainty, a fear that whispers at the edges of the sandstorm, the fear of machines making decisions of life and death on the battlefield. According to the International Committee of the Red Cross, artificial intelligence and machine learning systems could have profound implications for the role of humans in armed conflict, particularly with regard to the increasing autonomy of weapon systems and other unmanned systems, new forms of cyber and information warfare, and, more broadly, the nature of decision-making. Artificial intelligence is no longer science fiction; it is a presence in the war-torn landscapes of Africa.

Several African countries, including Zimbabwe, Eswatini, Angola, and Mozambique, have invoked national security to justify the deployment of surveillance technologies. The difficulty with mass surveillance is that it creates a climate of fear in which citizens are presumed guilty until proven innocent, and it alters the balance of power between a state and its citizens. In Nigeria, 2.2 billion naira was allocated in the budget of 2018 for a social media monitoring capability, alongside instructions to the military to watch for anti-government content online. In Libya, autonomous lethal weapons systems have reportedly been used in fighting. In Zimbabwe, a military-driven national facial recognition scheme has raised concerns over the technology’s alleged use as a surveillance tool. The draft continental policy of the African Union does not explicitly address the use of such technology by African governments for national security purposes, although it acknowledges that hazardous risks may arise.

MAIN CORPUS

Potential benefits in African militaries

In the field of peace and security, these technologies can enable more effective conflict analysis and early warning. They can support peacemaking and mediation by addressing information asymmetry. They can also enable state institutions to enhance their capacity for enforcing law and order and for countering criminality, thereby contributing to the security of citizens. Surveillance and policing platforms are deployed for tracking organised criminal networks and for responding to or preventing the activities of terrorist or insurgent groups.

The Nigerian Navy is adopting such technologies, which are increasingly used in ship construction and maintenance, as maritime operations become more complex. The Chief of the Naval Staff set this out during the presentation of a paper by naval participants at the National Defence College, on artificial intelligence and ship maintenance as a strategic option for the Nigerian Navy by 2035, stating that the service must continue to adopt and integrate these technologies in order to maintain a competitive edge during operations.

In light of the need for guidelines and legal frameworks that promote transparency, accountability, and compliance with human rights in the adoption and use of such technologies, at least seven African countries, namely Benin, Egypt, Ghana, Mauritius, Rwanda, Senegal, and Tunisia, have developed national programmes. While the adoption of such regimes at national level is important, the governance and regulatory challenges posed by these technologies exceed the capacity of individual states. One area where they are already in use, and will undoubtedly spread with unpredictable benefits and possibly much harm, is in the many wars and conflicts across the continent.

Risks and ethical concerns

These technologies also carry negative aspects, some of them of particular concern for Africa. As a general-purpose technology, artificial intelligence is susceptible to being used for harmful ends. There are increasing concerns associated with generative systems in relation to disinformation, cybersecurity threats, hate speech targeting women and minorities, and the incitement of violence in times of crisis and conflict. Synthetic voice and image material has been used to impersonate political figures and to propagate false information during elections in Nigeria and during the civil war in Sudan.

As Nate Allen of the Africa Center for Strategic Studies observes, artificial intelligence is already built into the technology used for surveillance platforms and satellite image analysis, and one application is the identification of movement such as tyre tracks. He notes that an autonomous fighter aircraft is not what is required, but rather an autonomous drone that can distinguish a tank from a car, and that what will work best in Africa is whatever is low cost and relatively easy to use. Drones have already been widely used in conflicts across the continent.

Unprecedented capabilities

Such systems, drawing on probabilistic reasoning and advanced algorithms, can compress the observe, orient, decide and act cycle to near-instantaneous speeds, creating scenarios in which machine-led interactions far outpace human cognition. This translates into greater battlefield awareness through the fusion of multi-source sensor data, enabling target identification and planning at scales beyond human capacity. It also facilitates a shift towards mass over quality, allowing the deployment of expendable autonomous systems, from drones to lethal autonomous weapons, that operate in hazardous environments with endurance and speed beyond human limits, reducing risk to personnel and capital investment.

Furthermore, these technologies strengthen both offensive and defensive capabilities in cyber operations, providing tools for more potent attacks and for continuously evolving countermeasures, and thereby becoming a strategic advantage that fuels a global arms race and alters power equations. This technological development promises a future in which wars are fought with enhanced speed, scale, and autonomy, demanding a fundamental reassessment of deterrence and of military strength.

International humanitarian law

International humanitarian law is at risk of being outpaced by the advance of these technologies. Who is accountable for the actions of an autonomous weapon system? How can respect for proportionality and distinction, principles enshrined in that body of law, be ensured when the decision belongs to an algorithm? The legal frameworks governing warfare have not caught up with this reality, leaving a dangerous vacuum.

There is some jurisprudence on the question, although it is scarce. In August 2020 the case of R (Bridges) v Chief Constable of South Wales Police was the first challenge in the United Kingdom to the use of such technology invoking human rights law. South Wales Police was trialling live automated facial recognition, comparing images of people attending public events with images held on a database. The Court of Appeal found that there was no proper basis in law for its use, and that the deployment did not satisfy the requirements of data protection legislation and of the public sector equality duty. The case halted the force’s use of the technology while leaving open the possibility of its reintroduction on a proper legal footing.

Some have raised the prospect of swarms of armed autonomous drones that are intended to distinguish combatants from civilians but that in practice may fail to do so. Others have suggested that these technologies could be used to anticipate conflict and civil unrest, allowing early intervention that preserves peace or mitigates violence. The Economic Community of West African States is understood to be working towards incorporating such methods into predictive models using information gathered from its members.

The advent of artificial intelligence, while reshaping the mechanics of warfare, also holds a nascent power to reduce war’s inherent uncertainty. Its capacity for rapid sensor fusion and probabilistic reasoning promises to reduce the fog of war, providing decision makers with greater situational awareness and potentially leading to more informed choices. This enhanced informational landscape could mitigate uncertainty about intentions and outcomes, a fundamental driver of conflict, and could prompt a convergence in expectations among international actors. Its potential to strengthen cyber defences could tip the long-term balance towards resilience and protection, eroding the structural advantages of attackers and possibly deterring aggression. The ongoing global discourse on meaningful human control over autonomous weapon systems, and the emphasis on ethical guidance, reflect a collective effort to ensure that human judgement, empathy, and moral consideration remain paramount, in the aspiration that the power of the technology may help to constrain the violence it could otherwise amplify.

SUMMARY *

The article’s proposal is to move the question from prohibition to design. On the author’s account, the debate over autonomous weapons has absorbed the available attention while the technologies that will actually shape African security, cheap drones, satellite analysis, facial recognition and social media monitoring, arrive without any comparable scrutiny, and several are already in use against populations rather than in armed conflict. The peace switch the author proposes is the claim that the same capabilities that permit targeting also permit early warning, analysis and mediation support, and that a governance framework built only to prohibit will regulate neither. The continental dimension gives this its force: the author observes that regulation exceeds the capacity of individual states, that the African Union lacks the instruments to legislate, and that responsibility will therefore fall back on national frameworks that mostly do not yet exist.

CONCLUSION

South Africa now stands as a continental pioneer, having inaugurated a defence-focused artificial intelligence hub within the Military Academy as one of the latest arms of the AI Institute of South Africa. This development is not an isolated step; it marks a shift in how African states are beginning to combine questions of war, peace, and machine logic. By contrast, the African Union remains structurally constrained in its ability to harmonise policy, to legislate enforceable standards, or to govern the military applications of emerging technologies, and its ambitions in this field remain aspirational. Even if the draft continental strategy secures parliamentary endorsement, it will still fall to sovereign states to give effect to the vision through national frameworks.

The crossroads is technological, philosophical, moral, and historical. Will Africa become a canvas for experimental militarised intelligence, haunted by the ghosts in the machine, or will it reimagine itself as a global leader in forging responsible, peace-centred governance of these technologies? These questions are not rhetorical flourishes. They are blueprints awaiting architects.

We may not yet possess definitive answers, but raising such questions is a beginning. It is an invitation to policymakers, to scholars, to African technologists and jurists, to create together the principles, processes, and pathways to peace that we do not yet have. In interrogating the uncertain, we prepare. The future of war and peace in Africa may well be written in algorithms, but the moral code must be authored by human hands. Africa’s technological awakening must not only be intelligent; it must be wise.

POTENTIAL SOLUTION *

The article’s central move, from prohibition to governance, is the right one, and the reason it is right can be stated more sharply than the article states it. The negotiations at Geneva on lethal autonomous weapons have run for more than a decade under a convention that operates by consensus, which means that any single state can prevent an outcome, and several with the relevant capabilities have done so. A prohibition that will not be agreed is not a policy. Meanwhile the systems that are actually deployed are not the ones under negotiation, and the article’s own examples establish this: facial recognition, social media monitoring, satellite analysis and low-cost armed drones are in use now, are procured commercially, and fall outside every instrument being debated.

That points to what a governance framework would have to cover, and it is broader than the weapons question. Export and procurement controls determine what enters a country and on what conditions. Data protection law determines what may be collected about whom. Judicial authorisation determines whether a surveillance capability may be turned on a population. Procurement transparency determines whether a system’s error rates are known before it is bought. None of these is a matter of arms control, and each is within the competence of a national legislature. The article’s observation that responsibility will fall to states is therefore not a counsel of despair but a description of where the work lies.

Two clarifications strengthen the legal analysis. First, the existing framework is not as empty as the metaphor of a vacuum suggests. The obligation under Additional Protocol I to review new weapons and means of warfare for compliance already applies to autonomous systems, and it is binding on states parties whether or not any new instrument is concluded; the difficulty is that very few states conduct such reviews rigorously and fewer publish them. Establishing a review capacity is achievable, is cheap relative to procurement, and would do more in the near term than the outcome of any negotiation. Second, the case cited from the United Kingdom is a stronger authority than the article suggests, because the ground on which it succeeded is transferable: the deployment failed not because the technology was inaccurate but because the discretion as to who went on the database and where the system was used was left to individual officers without a legal framework governing it. That is a rule of law argument, and it is available in any jurisdiction with a constitutional right to privacy, which most African constitutions contain.

Two further observations for this journal’s readers. The article’s claim that these technologies can support mediation and early warning is the least developed part of it and the most relevant to this field, and the record supports a modest version: automated monitoring of hate speech, satellite detection of displacement and crop failure, and the analysis of communications for signals of mobilisation have all contributed to warning systems, and each shares the weakness identified in every review of prevention, namely that the failures have concerned political will rather than information. A better detector attached to the same decision-making produces earlier reports of atrocities that are not prevented. And the dual-use point cuts both ways: the same platform that maps population movement for humanitarian planning maps it for targeting, which means the governance question is not what a system can do but who controls it and under what authorisation.

Finally, the article’s framing of Africa’s position deserves to be taken as more than rhetoric. The continent is currently a market rather than a producer, which means that the terms on which these systems arrive are set elsewhere, and that the accountability for their failures is correspondingly difficult to locate. The most consequential thing the states concerned could do is not to draft a declaration but to establish jointly the technical capacity to evaluate what they are being sold, because a purchaser who can assess a system is a purchaser who can impose conditions on it. That is a modest institutional proposal, and it is the one on which everything else in this article depends.

* Added by the WMO Editorial Team

REFERENCES

Anyemedu, D. Digital Surveillance in Violation of Human Rights and Data Justice. Accessed 26 July 2024.

Davison, N. (2017) Autonomous Weapon Systems under International Humanitarian Law. In Perspectives on Lethal Autonomous Weapon Systems, United Nations Office for Disarmament Affairs Occasional Papers No. 30, November 2017.

Hernandez, J. (2021) A Military Drone with a Mind of Its Own Was Used in Combat, UN Says. NPR, 1 June 2021.

International Committee of the Red Cross. Artificial Intelligence and Machine Learning in Armed Conflict: A Human-Centred Approach. Accessed 27 July 2024.

Mare, A., and A. Munoriyarwa (2023) Digital Surveillance in Southern Africa. Palgrave Macmillan, Cham.

Nigerian Navy to Harness Artificial Intelligence to Strengthen Operations. Accessed 25 July 2024.

R (Bridges) v Chief Constable of South Wales Police [2020] EWCA Civ 1058.

Equality Act 2010, section 149(1).

Tsani, A. (2024) Africa’s Push to Regulate AI Starts Now. MIT Technology Review, 15 March 2024.

Supplementary references added by the WMO Editorial Team

Protocol Additional to the Geneva Conventions of 12 August 1949, and relating to the Protection of Victims of International Armed Conflicts (Protocol I), 1977, Articles 35, 36, 48, 51 and 57.

Convention on Prohibitions or Restrictions on the Use of Certain Conventional Weapons, 1980, and the reports of the Group of Governmental Experts on Lethal Autonomous Weapons Systems, Geneva.

United Nations General Assembly Resolution 78/241 (2023) on lethal autonomous weapons systems.

African Union Continental Artificial Intelligence Strategy, adopted 2024, and the Malabo Convention on Cyber Security and Personal Data Protection, 2014.

African Charter on Human and Peoples’ Rights, 1981, and the Declaration of Principles on Freedom of Expression and Access to Information in Africa, 2019.

International Committee of the Red Cross (2021). ICRC Position on Autonomous Weapon Systems, Geneva.

Boulanin, V., & Verbruggen, M. (2017). Mapping the Development of Autonomy in Weapon Systems. Stockholm International Peace Research Institute, Stockholm.

Allen, N. (2024). Advancing Artificial Intelligence Governance in Africa. Africa Center for Strategic Studies, Washington D.C.

Buolamwini, J., & Gebru, T. (2018). Gender Shades: Intersectional Accuracy Disparities in Commercial Gender Classification. Proceedings of Machine Learning Research, 81, 1 to 15.