As states race to develop autonomous weapons systems, the effort to agree binding international rules on machine decisions over life and death has reached a critical moment.

In this post, Eberechukwu Owuamanam, Jesuit scholastic and humanitarian practitioner, argues that Pope Leo XIV’s call to “disarm” artificial intelligence (AI) should be taken literally when it comes to such weapons. Writing from a humanitarian perspective shaped by work in the Global South, he contends that no autonomous system should ever select and attack a human being without meaningful human control, warns that automated force will fall hardest on those least able to build or resist it, and calls on states to conclude a binding treaty prohibiting autonomous weapons designed to target people and strictly limiting the rest.

When Pope Leo XIV presented his first encyclical at the Vatican in May 2026, he reserved his sharpest words for the growing role of artificial intelligence in warfare. He warned that delegating decisions over the use of force to machines hollows out the very basis on which any war could be justified, arguing that the just-war tradition must be rethought in light of the profound changes in the means and methods of warfare. A papal text on AI is not a document for Catholics alone; it is an intervention in a debate the humanitarian world is already grappling with. And on this point, at least, I want to say plainly where I stand.

I have come to be convinced, through years of humanitarian work in conflict-affected communities in Nigeria, that artificial intelligence must be disarmed, and that the phrase deserves to be taken at its most literal. Human life is precious. The decision to end a human life must never be delegated to an autonomous or algorithmic machine; it requires a conscience. One could argue about accuracy or reliability; however, we must clearly understand who we are as human beings and what we owe others when their lives hang in the balance. I write this from a Global South vantage point, where such voices are often discussed but barely listened to, or even heard.

The conviction that AI should be disarmed is not entirely new. In 2023, ICRC President Mirjana Spoljarić joined the UN Secretary-General in a joint appeal urging states to negotiate and to conclude by 2026 a new legally binding instrument setting clear prohibitions and restrictions on autonomous weapon systems. That deadline is now. The question is whether the world will meet it, and on whose terms.

A line we must not cross

An autonomous weapon, stripped of euphemism, conscience, soul and spirit, is a machine that selects and attacks targets on the basis of sensor processing rather than a human conscience: the sober, unflinching awareness of what it costs to end a life. A human may authorize the system’s deployment, but no one, at the moment a life is taken, holds that death in conscience and chooses it. Hence, when such a system is pointed at human beings, one or thousands of human lives end without any sobriety or realization of the weight of the decision. The human who authorized it is never made to look the killing in the eye.

I do not believe this is a matter on which morally serious people can simply agree to disagree. Talk of “thresholds” and “acceptable risk” can obscure what is actually being decided. For me, this is a moral line. The UN Secretary-General has said as much, describing the autonomous targeting of humans by machines as “a moral line that must not be crossed”, and reporting that many states consider the delegation of the decision to take a human life to a machine simply unethical.

This conviction is not secondary to the ICRC’s own position. In its October 2025 position paper on autonomous weapons systems, the ICRC describes AWS as systems that, once activated, can select and engage targets without further human intervention, and states that new, legally binding rules are urgently needed, including prohibitions on autonomous weapons that pose an unacceptable risk of violating international humanitarian law, including those that are unpredictable, alongside strict regulation of the rest. The encyclical names the same wound from a different direction: the reduction of a person to a means, a thing to be processed rather than someone to be reckoned with. A weapon that kills without a human in the decision is the purest expression of that reduction.

In Libya in 2020, a loitering munition was reported to have attacked retreating fighters without a human authorizing the specific strike, widely cited as the first use of a lethal autonomous weapon against human beings. The threshold has already been stepped over once. The task now is to ensure it is not normalized.

The injustice the technology will entrench

Here I must say something that the comfortable center of this debate tends to leave unspoken, and that I cannot, as an African and a humanitarian, leave unsaid.

The capacity to build autonomous weapons is not evenly distributed across humanity. In 2025, the United States, China, and Russia together accounted for just over half of all military spending worldwide, and within those states, the capacity concentrates further among a handful of firms that decide what gets built and how fast. The brutal first-mover sense is clear and unforgiving. Small advantages in speed and lethality compound, and the gap between those who can build these systems and those who cannot does not narrow. It widens.

It is also worth asking who profits from this. Investment in autonomous weapons start-ups topped 146 billion US dollars in just the first five months of 2026, already beyond the previous full-year record, on margins several times those of conventional defense contractors. A market with its own reasons to ensure these systems are built, sold, and used, and that reason has nothing to do with protecting anyone.

Consider a weapon that can kill without risking the life of any soldier on the side that owns it. The moral sobriety of going to war is lost, but only for that side. The people most likely to be on the receiving end are those without the means to build, deter, or answer it, and that gap will not hold for long. Armed drones have already spread from fewer than ten countries in 2010 to more than forty today, and over sixty-five non-state armed groups now hold some drone capability. Autonomous weapons will follow the same path: a head start for the powerful, then a far less governable free-for-all.

That the gap may close is not a reason for less urgency. It is a reason for more. The same states building these weapons have, for over a decade, used the consensus rule of the Geneva disarmament process to block negotiations that might restrain them. The powerful are building the weapons and guarding the exit.

When the Maxim gun met spear and shield in the First Matabele War of 1893, and across the Scramble for Africa, technological asymmetry did not produce cleaner battles. It produced massacres, and it underwrote a century of domination. Autonomous weapons are the digital descendant of that same order, which is why a growing body of scholarship now speaks of digital colonialism and algorithmic coloniality. Delegations from the Global South have warned bluntly that AI-enabled weapons are likely to be tested on their populations. Africa, dependent for its arms on the very states that build these systems, is widely judged a likely theatre for them. What is at stake is not only a humanitarian risk but a durable new subjugation: a world where the resourceful can kill the resourceless at no cost to themselves, by machine, and without end. That is a re-colonization of force, and it would harden global inequality into something close to a caste of nations.

At the same time, the Caribbean through the CARICOM Declaration, Latin America through the Belén Communiqué, and West Africa through the ECOWAS Freetown Communiqué have all called for a binding treaty, and groups of smaller states have tabled their own drafts. The UN General Assembly has adopted a resolution on autonomous weapons three years running, by 166 votes to 3 in December 2024 and again, 164 to 6, in December 2025. More than 120 states are on record for a treaty. The majority of humankind is asking for restraint, and that voice must be answered.

A weapon the law has refused before

We have done this before, and we should learn from how. International law has, on rare occasions, examined a weapon and banned it. The prohibition of poison or poisoned weapons is a long-standing rule of customary international humanitarian law already recognized in the Lieber Code and the Hague Regulations reaffirmed as a binding custom by the International Court of Justice. Poison was banned not because it sometimes hits the wrong target, but because no one who uses it can predict or contain who it kills. That is the test autonomous weapons fail. A weapon that selects its own targets and acts in ways its operators cannot foresee belongs in the same category as the unaccountable, and deserves the same answer: not regulation alone, but prohibition.

The risks are now on formal record. The General Assembly’s most recent resolution warns of an emerging arms race, lowered thresholds for the use of force, miscalculation and rapid escalation, and proliferation to unauthorized recipients and non-state actors. Each is a grave danger on its own. Together, systems built to act faster than any human can review, fielded first by the powerful and later by whoever can acquire them, describe a path along which humanity loses oversight of its own instruments of violence.

There is a closer precedent than nuclear weapons. In 1995, states banned blinding laser weapons under the Convention on Certain Conventional Weapons before their use had spread, proof that preventive regulation can work when it acts before a weapon is built into arsenals and doctrine. Autonomous weapons sit at that same threshold now: used in isolated cases, not yet everywhere.

The Seventh Review Conference of the CCW in November 2026 is the moment to open negotiations on a new protocol. Disarming autonomous weapons will take the same coalition the blinding-laser ban required: states and civil society insisting the technology be governed before its use becomes ordinary, rather than bargaining over limits on a weapon already built into every arsenal. A binding instrument is not an obstacle to national security. It is the only way to stay ahead of the danger.

AI is not the enemy

This post is not a polemic against artificial intelligence. The encyclical itself is careful in asserting that technology is not, in itself, antagonistic to humanity, nor is it evil, though it is, the Pope insists, never neutral, taking on the character of those who design, fund, and deploy it.

Used well, AI can help build a common world. It can widen access to education where teachers are few, open trade to those long shut out of global markets, strengthen governance, speed up medicine, and make humanitarian work faster and fairer. These are among the most hopeful tools of our time, and the Global South has as much claim on them as anyone. To disarm AI, in the encyclical’s sense, is not to reject it. It is to free it from the logic of armed competition and return it to human purposes.

The plurality the encyclical calls for

If we are to build AI that serves rather than dominates, the harder question is whose understanding of the human will guide it. The encyclical declines the comfortable assumption that simply “aligning” AI with human values is enough, and asks: on whose values, and aligned to which picture of the person? A morality settled by a few, it suggests, is not yet a shared morality at all. Furthermore, the Pope proposes that disarming AI means restoring it to “the plurality of human cultures and ways of life,” which are critical and must be respected.

This is an invitation, and the traditions that could answer it are many: Asian relational philosophies, the Indigenous philosophies of the Americas, and the African relational traditions within which I work. Ubuntu, the Igbo principle of Igwebuike, and a framework I am developing called Ijeluwa, which holds that a person is not an isolated individual who later enters relationships but a being constituted through them, are among the many voices the encyclical’s plurality invites. Their relevance to the question of AI reaches beyond the battlefield.

The same reductive assumptions that let a machine end a life also let a database reduce a refugee to a datum, as I argued in an earlier reflection on this blog, where humanitarian systems too often help individuals while harming persons. Whether the question is a weapon that targets a person or a biometric registration system that processes one, a relational tradition asks the same thing: does this system see the person as the web of relationships they are, or only as a unit to be processed? That is a test worth building into how we govern AI, one of many the world’s traditions could help us write.

What I am calling for

So let me end where my conviction begins. I am calling, as a humanitarian and from where I stand, for three things.

First, that states open negotiations at November’s CCW Review Conference on the binding instrument the UN Secretary-General and the ICRC’s own President have urged: a prohibition on autonomous weapons whose effects cannot be reliably predicted or controlled, and on any system designed to target human beings, with strict regulation of the rest. The autonomous killing of a human being should be as unthinkable, and as unlawful, as the use of poison, a line the law has held for over a century, whoever holds the weapon.

Second, that we name and resist the inequality this technology threatens to entrench, and treat the leadership of the Global South not as an afterthought but as central to this debate, recognizing this is not a single voting bloc with one position, but dozens of states, regional declarations, and civil society coalitions that have done the harder work of building consensus across real differences, because those most exposed to autonomous force have also done the most to demand it be restrained.

Third, that we take up the encyclical’s invitation and build the AI that will shape education, trade, governance, and humanitarian action in conversation with the plurality of human traditions, so that the question of who a person is, and what is owed to them, is answered before a machine is permitted to act on the answer.

Pope Leo XIV closes his encyclical with the image of a construction site: a wall being raised with a section waiting for each of us. The section waiting for the humanitarian world is this one. Human life is precious, and the Church has long insisted that no one rightfully possesses the authority to end it. That is precisely why taking a life must never be delegated to an algorithm beyond all moral reckoning. From where I stand, that is what disarmament means, and it is the most urgent thing the people I have spent my life among are owed.

Related articles

• Laura Brunn, Reinventing the wheel? Three lessons that the AWS debate can learn from existing arms control agreements, June 13, 2024
• Alexander Blanchard, The road less travelled: ethics in the international regulatory debate on autonomous weapon systems, April 25, 2024
• Ingvild Bode, Falling under the radar: the problem of algorithmic bias and military applications of AI, March 14, 2024
• Ruben Stewart and Georgia Hinds, Algorithms of war: The use of artificial intelligence in decision making in armed conflict, October 24, 2023