Lethal autonomous weapons: can the decision to kill be entrusted to a machine?

My master’s research in human rights law examined the compliance of Western armed forces with international humanitarian law (IHL) in urban settings. Of all the technological transformations of warfare — drones, smart munitions, comprehensive battlefield surveillance — only one truly changes the nature of the legal problem: the one that removes the decision to kill from human hands. It is this question that I take up here.

1. What are we talking about?

Two families must be distinguished. Lethal weapons systems that integrate autonomy carry out certain functions automatically, but remain under human control: an operator validates, corrects, interrupts. Lethal autonomous weapons systems (LAWS), by contrast, are designed to make the decision to attack without any assessment of the situation by the chain of command once they are under way.

The gap is not one of degree, but of nature. It has been described as the “third revolution in warfare,” after gunpowder and nuclear weapons. But with one decisive difference: for the first two, the military retained the role of final decision over the use of force. With a fully autonomous LAWS, no human supervision remains possible at the moment of the strike.

Between total human control and total computer control lies a continuum — the ten levels of autonomy described as early as 1978 by Sheridan and Verplank range from “the human does everything” to “the computer acts if it decides it should be done, and informs the human only if it deems it useful.” The real question, therefore, is not “should there be autonomy?” but how far it is allowed to go, and where the human must remain in the loop.

2. Humanitarian law put to the test by the machine

IHL imposes three cardinal principles on any attack. A LAWS would have to be capable of respecting them — and that is where everything becomes complicated.

• Distinction. Target only military objectives, never civilians. A well-designed system can, in theory, recognise a tank. But in an asymmetric confrontation, in a dense urban area, with reduced visibility, how does a machine distinguish a combatant from a civilian — a man laying down a weapon from a man picking one up?
• Proportionality. Weigh the anticipated military advantage against the foreseeable civilian harm. This is an ethical and human judgement, contextual, ill-suited to a programmable rule.
• Precaution. The effects of the attack must be foreseeable, so that it can be adjusted or called off. Yet unpredictability is precisely the criticism levelled at the most autonomous systems.

IHL, it bears recalling, is a body of law by and for the human being: it codifies the effects of actions taken by persons, in order to alleviate human suffering. A weapon that, once launched, rules out any human intervention runs into head-on tension with this logic.

3. The black hole of accountability

Then comes the question that the law does not know how to resolve: who is responsible when an autonomous machine commits the irreparable?

Traditionally, one traces responsibility back to the person who fired or to the superior who ordered the attack. But a LAWS has neither free will nor awareness of the scope of its acts. Criminal responsibility presupposes two elements — intent (mens rea) and the attributable material act (actus reus). The total absence of intent makes both impossible to establish. One does not convict a machine; one does not seek reparation from it. Human Rights Watch summed up this deadlock in a phrase: “mind the gap” — beware the accountability void. This void is not a technical detail: it is a breach in the very promise of justice for victims.

4. An international law at a standstill

Faced with this, the international community has not remained inert — but neither has it settled the matter.

Since 2014, at France’s initiative, LAWS have been discussed within the framework of the UN Convention on Certain Conventional Weapons (CCW); a Group of Governmental Experts has been working on the subject since 2017. The European Parliament came out in favour of a ban as early as 2018. In 2019, a large share of UN member states described these weapons as “morally repugnant,” in the words of António Guterres.

But no binding convention has emerged, for a simple reason: several major military powers — the United States, Israel, the United Kingdom, Australia, Russia — oppose any legal regulation, having invested massively in these technologies.

The avenues nonetheless exist. The 2019 Franco-German declaration puts forward eleven guiding principles: subjecting these systems to IHL, keeping the decision to use them under human responsibility, reviewing the lawfulness of the weapons from the design stage. In 2021, the ICRC published clear recommendations: legally binding rules, a prohibition of unpredictable systems and of those designed to target human beings, and strict regulation of the rest (types of targets, duration, geographic scope). At the heart of all this, one notion: meaningful human control.

In 2023, Guterres called on states to produce, by 2026, a legally binding instrument prohibiting weapons that operate without human supervision and that do not comply with IHL. A General Assembly resolution, adopted by 164 states in October 2023, reaffirmed the urgency — but without binding force. A symbolic gesture, not a norm.

5. 2026: the moment of truth, or the fateful utopia?

Here we are. The deadline has arrived, and the obstacles that stood in the way yesterday are still standing: no shared definition of “autonomy,” and the reluctance of military powers to limit technologies in which they have staked everything. Without global consensus, genuine international regulation will remain — to borrow a phrase from my dissertation — a fateful utopia.

My conviction is simple, and it reaches beyond the battlefield alone. The line not to be crossed is not technological, it is moral: a human being must retain meaningful control over the decision to resort to force. This holds true for autonomous weapons; it is the same principle that runs, quietly, through the whole governance of artificial intelligence — health, justice, security — whenever there is a temptation to delegate to a machine a decision with irreversible consequences.

Technology can be an ally. It cannot be an alibi for evacuating human responsibility. Between operational efficiency and respect for the law, there is no choosing: it is the latter that sets the limit of the former.

Bibliography

• Ancelin, J. (2016). “Les systèmes d’armes létaux autonomes (SALA) : enjeux juridiques de l’émergence d’un moyen de combat déshumanisé.” La Revue des droits de l’homme.
• Barrier, F. (2018). “Les systèmes d’armes létaux autonomes (SALA) : vers une nouvelle course à l’armement ?.” Revue Défense Nationale, 810(5), p. 19.
• Bernard, V. (2012). “La science ne peut pas être placée au-dessus de ses conséquences.” Revue internationale de la Croix-Rouge, 94(2).
• Brunn, L. (2022). “Les systèmes d’armes autonomes : ce que le droit dit – et ne dit pas – sur le rôle de l’être humain dans l’usage de la force.” Droit et politiques humanitaires (ICRC blog).
• Casier, F. (2021). La réglementation des systèmes d’armes autonomes. Une réponse nécessaire aux enjeux éthiques, humanitaires et juridiques. Croix-Rouge de Belgique.
• Comité d’éthique de la défense (2021). Avis sur l’intégration de l’autonomie dans les systèmes d’armes létaux. Ministère des Armées, Paris.
• International Committee of the Red Cross (2021). ICRC recommendations on the regulation of autonomous weapons systems.
• International Committee of the Red Cross (2022). Q&A – What you need to know about autonomous weapons.
• Convention on Certain Conventional Weapons (CCW), Geneva, 10 October 1980 (entered into force on 2 December 1983).
• Cottier, D. (2022). Rapport sur l’émergence des systèmes d’armes létales autonomes et leur nécessaire appréhension par le droit européen des droits humains. Parliamentary Assembly of the Council of Europe, AS/Jur (2022) 30.
• International Court of Justice (1996). Advisory Opinion of 8 July 1996 on the Legality of the Threat or Use of Nuclear Weapons, ICJ Reports, § 86.
• Gayle, D. (2019). “UK, US and Russia among those opposing killer robot ban.” The Guardian.
• Human Rights Watch (2015). Mind the Gap: The Lack of Accountability for Killer Robots.
• Kellenberger, J. (2011). “Le droit international humanitaire et les nouvelles technologies.” 34th Round Table on Current Issues of International Humanitarian Law, San Remo.
• Krieg, A. & Rickli, J.-M. (2019). Surrogate Warfare: The Transformation of War in the Twenty-First Century. Georgetown University Press.
• McFarland, T. (2015). “Les facteurs permettant de déterminer les conséquences juridiques liées à l’utilisation de systèmes militaires toujours plus autonomes.” Revue internationale de la Croix-Rouge.
• Ministère de l’Europe et des Affaires étrangères (2019). 11 principes sur les systèmes d’armes létaux autonomes (Franco-German declaration, CCW).
• Missiroli, A. (2020). “Game of drones? Ou comment les nouvelles technologies influent sur la dissuasion, la défense et la sécurité.” NATO Review.
• United Nations, General Assembly – First Committee (2023). Lethal autonomous weapons systems (A/C.1/78/L.56), 78th session.
• European Parliament (2018). Resolution of 12 September 2018 on autonomous weapon systems. Strasbourg.
• Sheridan, T. B. & Verplank, W. L. (1978). Human and Computer Control of Undersea Teleoperators. Massachusetts Institute of Technology.