In his first encyclical, Magnifica Humanitas, Pope Leo XIV repeatedly emphasizes the need to “safeguard” the “primacy” of the human in the age of artificial intelligence (AI). According to Leo, humans are morally primary to so-called AI “agents” in at least two ways: first, in their dignity and intrinsic worth, and second, in their nature as responsible beings. When he turns to analysis of the potential role of AI in modern warfare, then, Leo focuses on the impact of war on human beings and the question of how responsibility can be assigned for those impacts. In conversation with Leo and feminist philosophers, I argue that the evasion of responsibility for harm is a persistent problem in war. While AI certainly opens up a new pathway for such evasion, the refusal of responsibility is a moral problem rather than a technological one.
Weakness, Error, and Responsibility
Human responsibility can take many forms; it includes responding to others as persons, taking moral and other forms of accountability for past actions, and intentionally learning over time to more reliably enact solidarity with fellow human beings. Pope Leo argues that mutual responsibility is an inescapable limitation of human life, yet “it is precisely within our limitations that the following find a place: compassion, as well as a sincere concern for the needs of others; a generosity that can emerge even in the midst of darkness and failure; spiritual experience and the worship of God” (para. 119). Leo identifies moral weakness as humans’ biggest limitation—and therefore, one of our most significant means of developing moral responsibility (para. 121).
Though Leo’s reasons for emphasizing the finitude of human beings are theological, his line of argument here closely tracks with those of feminist ethicists who have insisted on the thoroughly interdependent nature of human beings. Like Leo, feminist moral philosophers have argued that these interrelationships simultaneously limit human action and generate our ability to act in responsible and empathetic ways. Though our moral standing is vulnerable to the effects of moral luck and our general inability to exert total control over the shape and outcomes of our actions, we cultivate virtue when we take responsibility in spite of these realities—when we accept that, in Margaret Urban Walker’s words, “responsibilities outrun control” (26).
One of the most powerful impulses in the adoption and incorporation of AI into warfare, as well as other aspects of human life, is the desire to eliminate weakness and error. But Leo rightly emphasizes that with human weakness and limitation come responsibility and the opportunity for growth. He denies that machines share these qualities:
So-called artificial intelligences do not undergo experiences, do not possess a body, do not feel joy or pain, do not mature through relationships and do not know from within what love, work, friendship or responsibility mean. Nor do they have a moral conscience, since they do not judge good and evil, grasp the ultimate meaning of situations, or bear responsibility for consequences. They may imitate language, behavior and analytical skills, or even simulate empathy and understanding, but they do not understand what they produce, for they lack the affective, relational and spiritual perspective through which human beings grow in wisdom. Even when these tools are described as capable of ‘learning,’ their way of doing so is different from that of a human person. It is not the experience of those who allow themselves to be shaped by life and grow over time through choices, mistakes, forgiveness and fidelity. Rather, it is a form of statistical adaptation based on data and feedback, which can be very effective, but does not imply inner growth”(para. 99).
As he puts the point more pithily elsewhere, “For an algorithm, an error is a flaw to be corrected; for a person, however, an error can be a catalyst for profound change” (para. 128). Throughout the text, Leo returns to the significance of responsibility, convincingly suggesting that it constitutes the central moral question for our consideration of the potential role of AI in war.[1]
Evading Responsibility
Leo connects the loss of responsibility to what he calls the “normalization” of war: “Here the question is not merely the efficiency of new tools, but also the risk that technology, detached from ethics and responsibility, will render decisions about life and death more rapid and impersonal, and will present the use of force as an immediate and viable option” (para. 182). Notably, though Leo does not use the phrase “presumption against war” here, the content of his argument implies that he agrees with those, like the U.S. Conference of Catholic Bishops and James Childress, who argue that the decision to go to war requires overriding a prima facie duty not to injure or kill other persons. As Aline Kalbian argues, this presumption is a reflection of what feminists have identified as the essentially relational nature of human autonomy and our resulting mutual vulnerability and mutual responsibility. When decisions about war are not made by flesh-and-blood, feeling, living persons, Leo implies, those decisions are less likely to take into account the effects of violence on other living persons, rendering the choice to use force an easier one. In a headline-garnering sentence from the encyclical, he concludes that under these conditions, “just war theory, which has all too often been used to justify any kind of war, is now outdated” (para. 192).
I share Pope Leo’s concern regarding responsibility and normalized violence, but I would argue that the integration of AI into war, far from introducing this issue, continues a long-standing history of evading responsibility for violence. Indeed, the ways in which AI has been developed and deployed for the financial and political gain of its creators, but with little accountability for its impact on human well-being, bears a striking resemblance to how war has been conducted throughout human history. More recently, American leaders have repeatedly signaled an unwillingness to take responsibility for the perpetration of (or complicity in) violence that harms civilians. In 2019, President Trump pardoned a former Army officer convicted of ordering the murder of two civilians. Throughout the last fifteen months of his term—a period during which over 40,000 Palestinians died in the war in Gaza—President Biden repeatedly defended and materially supported the actions of the Israeli military, while publicly questioning the accuracy of the death toll counts provided by the Gazan health ministry. These trends have accelerated during the second Trump administration. Secretary of War Pete Hegseth dismantled the military’s Civilian Harm Mitigation and Response office and has stated that he envisions a military focused on “maximum lethality” rather than “tepid legality.” This rejection of legality has been publicly enacted in Hegseth’s campaign of boat bombings in the Pacific and Caribbean, which have killed over 240 people. The laws of war do not permit targeting civilians suspected of criminal activity (and for good reason; evidence suggests that many of those killed have been not drug traffickers but fishers and human trafficking victims).
In another example, President Trump continues to deny—in the face of conclusive evidence to the contrary—that the United States was responsible for the February airstrike on an Iranian school that killed over 175 people, mostly children. In the immediate aftermath of this strike, especially as reporting emerged confirming that AI programs had been used to assist in generating target lists in Iran, there was much speculation that this strike was an early example of the kinds of catastrophic mistakes of which we might expect many more as a result of military reliance on AI. Subsequent investigation has suggested, however, that the strike was instead caused by a failure to update imagery of the target site for over seven years—even after an analyst warned that it appeared a school had been built there. This was a human failure, with devastating results. It is also one for which no responsibility has been taken, even in the most minimal form of an acknowledgment by the U.S. government that it was our military that perpetrated the strike. I dwell on this example first, because it reminds us that the violence of war is not abstract but real and deadly, and second, because it so clearly demonstrates that the evasion of accountability for the infliction of harm is not tied to any particular technology.
Revisiting Responsibility and Just War
That said, within the existing culture of responsibility evasion that I’ve described, I do not doubt that Leo is right that AI use will further enable harmdoing without accountability. But I worry about the implications of suggesting that one of the most historically significant accountability mechanisms in the context of war—the tradition of just war thinking and the subsequent development of international humanitarian law, based in large part on just war criteria—has been rendered irrelevant. This claim should not be taken to suggest that just war reasoning ought not to be subject to vigorous criticism. To the contrary, debate and disagreement have long been part of the just war tradition and continue to be so. Given that many of the important moral concerns Leo raises about contemporary war can in fact be rendered in the language of just war criteria including just cause, last resort, proportionality, and discrimination (see, for example, paras. 182–83, 189, 197, and 203), I wonder if his point might be more accurately described not as the claim that just war reasoning is outdated, but instead as an argument for the expansion of responsibility for violence.
When Augustine of Hippo—Pope Leo’s most significant spiritual influence, and the man often described as the founder of Christian just war reasoning—wrote of decisions to go to war, he made clear the centrality of humanity. Though he believed that past just wars had led to good consequences, in the City of God Augustine urged his readers to “Consider the scale of those wars, with all that slaughter of human beings, all the human blood that was shed!” Thus, he argued, while waging a just war may be necessary, one who “remembers that he is a human being” will “lament” that such necessity exists. “[A] man who experiences such evils, or even thinks about them, without heartfelt grief, is assuredly in a far more pitiable condition, if he thinks himself happy simply because he has lost all human feeling” (XIX.9). From a feminist perspective, Kalbian similarly writes that “Agents in a dilemmatic situation are never fully relieved of the traces of guilt or sorrow for what they have done, and that is a good thing. It serves as a reminder of our shared vulnerability and woundedness as human subjects” (293). AI cannot replace these human feelings and their role in moral responsibility, and we are right to reject any moves to replace the responsible decision-making of humans with autonomous algorithms. But we must not fail to be vigilant against all the ways that humans have devised of avoiding responsibility without the assistance of any technology. What Leo reminds us of is the essential fact that to fail to center the human in decision-making is simultaneously to fail to center the humans who will bear any suffering and harm that issue from those decisions.
[1] Leo seems to consider our inability to hold AI accountable for its decisions a decisive strike against its use in warfare. Some might respond to him that while we currently lack legal mechanisms for such accountability, that doesn’t mean they could not be developed in the future. Indeed, in the context of domestic law, moves are already being made in this direction. For instance, investigations showed that the perpetrator of an April 2025 deadly mass shooting on my campus, Florida State University, consulted with ChatGPT in the weeks and moments leading up to the crime. The state of Florida has now opened a criminal investigation into its role in the crime, and several of the victims’ families have filed lawsuits against OpenAI, its parent company. Importantly, however, while these mechanisms may allow for responsibility in the form of determining liability to punishment or compensation, they fail to enable the other, more forward-looking aspects of responsibility Leo describes. Leo sometimes distinguishes between these two dimensions of responsibility by differentiating the terms “accountability” and “responsibility” in the text; see for example Pope Leo XIV, Magnifica Humanitas, 105. On ChatGPT and the FSU shooting: https://www.npr.org/2026/04/21/nx-s1-5793967/florida-openai-investigation-mass-shooting-fsu; https://www.pbs.org/newshour/nation/lawsuit-accuses-chatgpt-of-helping-gunman-plan-fsu-mass-shooting
