Jump to page content

Seventh Review Conference of the Convention on Certain Conventional Weapons - Position Paper on lethal autonomous weapons systems

  • 19.11.2026
    • Disarmament
Scroll to page content

Position Paper of the Holy See on lethal autonomous weapons systems

submitted to the Seventh Review Conference of the Convention on Certain Conventional Weapons (CCW)

Geneva, 16-20 November 2026

                Introduction

1.    Technological progress in the military, including the increasing autonomy of weapons due to artificial intelligence (AI), has profoundly altered the nature of warfare. However, machines can make only calculations, and moral judgement cannot be reduced to a calculation; it involves conscience, personal responsibility and recognizing others as people. Therefore, it is not acceptable to entrust irreversible decisions, such as those involving life and death, to artificial systems. AI makes war faster and more impersonal, “lowering the threshold for resorting to violence, transforming defense into threat prediction and thus reducing victims to data.”[1]

2.    When algorithms are entrusted with decisions that would traditionally have been based on human judgement, it distances political and military decision-makers from the human cost of conflicts, thus giving rise to an even more inhumane conduct in the immense tragedy that war is[2]. Such an approach to war leads to violence being considered as inevitable. It can spiral out without limit, and cause moral injury, while the sole purpose becomes the optimization of technological use.

3.    Surpassing cold, logical and utilitarian criteria, war requires preserving the consciousness of the fragility and misery of human nature. In this regard, one should not forget that the dynamics of reconciliation, the post-bellum period and peacebuilding also depend on how hostilities are conducted. Regrettably, the battlefields have turned into life testing grounds for ever more sophisticated and autonomous weapon systems, thus adding to the suffering and destruction caused by war. The “anthropomorphizing” of weapon systems, combined with the “dehumanization” of the enemy, calls into question the applicability and the relevance of legal frameworks that were painstakingly developed in the wake of past conflicts. These tragedies made humanity more sensitive to the suffering of others. The proliferation of uncontrolled autonomous systems and high-speed processes is now outpacing the ability to implement adequate safeguards in place to preserve fundamental humanitarian and ethical principles, leading to conflicts and disproportionate reactions. The resulting moral, legislative, and regulatory gap could have far-reaching consequences for humanity.

4.    Against this backdrop, it is worth recalling that the future of peace also depends on our collective decisions. The CCW preamble recognizes “the importance of pursuing every effort which may contribute to progress towards general and complete disarmament under strict and effective international control”. It is indeed necessary to continue improving protections, including through “the codification and progressive development of the rules of international law applicable in armed conflict”.

5.    The questions raised by lethal autonomous weapons systems (LAWS) have been examined extensively for over a decade now, with broad convergence towards the issues at stake emerging throughout the years, most recently, through the adoption of the final report of the Group of Governmental Experts of the High Contracting Parties (GGE) related to emerging technologies in the area of LAWS,[3] on 3 September 2026. The time has now come to transform this shared knowledge and consensus into a concrete, operational framework.

6.    In light of recent developments in AI, the Holy See wishes to reiterate some of the ethical and legal concerns that it has raised, proposing a way forward to address ethical and legal issues. This follows the invitation of Pope Leo XIV in his Encyclical Letter Magnifica Humanitas “to establish a shared framework — also at the international level — in order to curb the technological arms race and ensure robust protection for civilians and the infrastructures necessary for their survival.”[4]

II.              Human dignity and the ethical claim upon LAWS

7.    Throughout history, when ethical considerations have been given a lower priority, the human cost has often been exorbitant, with the innocent paying the highest price. However, history also demonstrates that humanity is capable of establishing institutions that protect our shared life, with a human-centered approach, thereby resisting the temptation to see the “other” only as an enemy or a statistical entity to be eliminated.

8.    Precisely because “every human person possesses an infinite dignity, inalienably grounded in his or her very being, which prevails in and beyond every circumstance, state, or situation the person may ever encounter”,[5] the human being cannot be reduced to meaningless and interchangeable strings of data. Personal dignity cannot depend on computational performance, military utility or algorithmic classification. Conscience, responsibility and the recognition of another person cannot be reduced to calculation[6]. The acknowledgment of the inalienable God-given human dignity as a gift preceding achievement[7] presupposes therefore that the human person remains the measure. Therefore, LAWS “which are capable of identifying and striking targets without direct human intervention, are a ‘cause for grave ethical concern’ because they lack the ‘unique human capacity for moral judgment and ethical decision-making.’”[8]

9.    Likewise, moral judgement cannot be programmed away. In no way can programming ethical rules be considered a solution that would allow responsibility to be transferred from human decision-makers to a machine. In effect, machines can only produce technical, algorithmic choices and can only operate in simulations of reality and of human behavior, without truly understanding or capturing all their incalculable dimensions. Reality, as such, can never be reduced to ‘representation’ or ‘simulation’ of itself. Such a simulation can only take account of the physical world and ignores non-physical elements, such as consciousness, subjective experience, autonomy, freedom, and the history of relationships. To decide whether an action (or omission) is legal and legitimate from an ethical point of view, it is also necessary to interpret contingent and specific contexts, which cannot be subsumed under universal formal rules, also because of the intrinsic limits of the computation process itself, nor a priori pre-established and pre-trained in algorithms. Moreover, the training environments are always limited and can never be sufficient to face random and complex war contexts. Hence, “no algorithm can make war morally acceptable”.[9] In this regard, expressions such as “artificial moral agents” are particularly concerning as they imply that machines could distinguish between right and wrong and do so with the same moral weight as human judgment[10]. Human beings must remain on moral command, not merely “in the loop”.

10. Despite its limitations, the progressive development of international humanitarian law (IHL) in the wake of deadly conflicts attests to the fact that amidst military operations, IHL seeks to limit human suffering caused by armed conflicts and mitigate their consequences. Accordingly, the principle of distinction must always be respected and the protection of civilian populations must remain a paramount consideration in the conduct of hostilities. In effect, any technology that facilitates attacks while disregarding human dignity and unnecessary suffering lowers the moral threshold of conflict. However, while compliance with IHL is indispensable, it does not exhaust ethical discernment, especially in the case of LAWS. There are various reasons for this.

11. Firstly, compliance with IHL does not, in itself, resolve the fundamental question of whether decisions regarding the use of lethal force can be delegated to autonomous systems.[11]

12. Secondly, there are behaviors involving LAWS that, although not explicitly prohibited by IHL, remain forbidden by the dictates of public conscience, spiritual values, experience and soldierly virtues. These are ultimately out of respect for the very same “humanity” that IHL seeks to preserve, as required by the Martens Clause.

13. Thirdly, even LAWS compliance to IHL requires a conscientious assessment. The recent GGE Report rightly affirms that “to uphold compliance with rules of international law applicable in armed conflict, including the principles and requirements of distinction, proportionality and precautions in attack, control and human judgement with regard to LAWS are needed”[12]. In effect, human judgement may not be satisfied merely by anticipating that a system’s effects will comply with IHL. In particular, LAWS would lack the ability to comprehend the gravity of causing injury or unnecessary suffering, conducting indiscriminate attacks, and launching attacks expected to cause excessive incidental civilian harm. These concepts, which are codified in IHL, require interpretation, good faith, and prudent judgement, all of which are uniquely irreplaceable human virtues. Therefore, classification based on behavioral patterns or inferred characteristics cannot realistically replace the full assessment required by IHL.

14. Fourthly, compliance with IHL implies the predictability and reliability of the weapons systems. Unforeseen interactions between algorithms of opposing factions could trigger a rapid and uncontrollable escalation. However, due to their autonomy and self-learning capabilities, the effects of LAWS may prove impossible to predict a priori. Furthermore, programming errors, data biases, or cyberattacks could lead to catastrophic targeting errors.[13] In such cases, LAWS could adopt behaviors that were neither anticipated nor authorized by humans. Likewise, due to their inherent unpredictability and the absence of human supervision, the use of swarms of LAWS, can have indiscriminate and unintended effects, posing a high risk to civilians and contradicting the fundamental principles of IHL. In fact, when changes to software, training or operating conditions may alter behavior, additional scrutiny is needed, in line with Additional Protocol I to the Geneva Conventions, whose article 36 requires scrutiny of new weapons, means or methods of warfare against applicable international law, including restrictions on their use in some or all circumstances. Such reviews should include independent expertise and realistic testing.

15. Whenever decision-making powers over the use of force are delegated to a weapons system the crucial action/responsibility nexus would be inevitably jeopardized, undermining the essential legal foundations on which international institutions are based, including the prevention of crimes and the prosecution of those who commit them. This highlights the critical issue of ensuring effective and self-aware human agency.

III.            The imperative of effective and self-aware human agency

16. In front of the pressing temptation to delegate decisions about life and death to machines, Pope Leo XIV has warned against “an unprecedented and destructive betrayal of the legal and philosophical principles of humanism that underlie and safeguard every civilization.”[14] Even reliability would not resolve the prior moral question of whether a lethal decision has been inappropriately delegated, because a consistently functioning system may still embody an unacceptable transfer of authority.

17. The Holy See welcomes the fact that High Contracting Parties clearly acknowledged that “IHL imposes obligations on States, parties to armed conflict and individuals, not on machines”[15], as responsibility is uniquely human and presupposes the ability to act in the exercise of freedom. The human person, whether a political authority or the military commander, can in no case be forced to accept the results of actions carried out by LAWS that he or she would not have consented to in conscience, or be prevented from questioning the suggested course of action. This would devoid political and military authority of its human content and significance.

18. Therefore, the decision to use lethal force cannot be delegated to opaque or automated processes, but must remain under effective, self-aware and responsible human control.[16] As a proposed operational application, authorisation should be informed by the actual context, the system’s limitations, the persons and objects potentially affected, and the foreseeable consequences of its use. These concerns may be made worse by the complexity and opacity of algorithmic processes, which can make their outputs difficult to understand, verify or challenge. Furthermore, authorised limits must be enforceable under foreseeable conditions, including degraded communications. If these limits cannot be maintained, the system should not be used in that configuration.

19. In all cases, technical assistance must remain subordinate to judgement for which identifiable persons can answer, which also requires preserving time for moral judgement. While AI tends to expedite decision-making processes, speed and efficiency should never be the supreme motivating force for the irreversible decisions taken to conduct an attack. Excessive reliance on automated assessments can undermine independent human judgment even when formal human oversight is in place. This is particularly relevant when automated recommendations arrive too quickly for responsible assessment. Interfaces, staffing and operating procedures must afford the competent human authority a genuine opportunity to question a recommendation, withhold force and intervene before harm becomes irreversible. Where the operational tempo renders such judgement impossible, nominal supervisory power cannot establish effective control.

20. Moreover, accountable decisions require an identifiable and verifiable chain of responsibility and the capacity to reconstruct decision-making.[17] In particular, those who design, train, authorize and employ technology must be held accountable for their decisions so that accountability is not transferred to “the machine”.[18] Moral responsibility, State responsibility, individual criminal responsibility and other forms of legal liability require their respective conditions to be established; none may simply be assigned to “the machine”. In this respect, practical implementation, investigation and remedies could be supported by documented roles, system version records and authorisations, operational logs and access for competent investigators.

IV.            The way forward – Upholding human dignity in the future

21. “The future of artificial intelligence is not pre-determined. As a technology developed by human beings, its future depends on the choices we make. We therefore have the responsibility to guide, change and, when necessary, reject its applications”[19]. The development of increasingly sophisticated weaponry, even when intended to reduce collateral damage, does not constitute a lasting solution. The Holy See is convinced that a legally binding agreement to address the questions raised by LAWS, subscribed to by all States, would advance security for all, help safeguard the integrity and the dignity of the human person, civilians and military alike. In the meantime, the Holy See reiterates its call for a moratorium on the development and use of LAWS, with a clear and specific prohibition on those that target human beings or whose use can be expected to cause accidental human casualties.[20]

22. While respect for IHL is fundamental, the Holy See also deems it essential not to lose sight of the primary and fundamental goal of cultivating peace and justice. As Pope Leo XIV affirmed, AI has “to be ‘disarmed’, freed from logics that turn it into an instrument of domination, exclusion and death”[21]. “Disarming technology” means recognizing that certain possibilities opened up by human ingenuity must be halted or subject to common constraints, because they have a direct impact on human dignity and the future of peace. To disarm does not mean to be against technological progress – quite the contrary. In the security domain, AI can contribute to the protection of civilians, early warning and crisis preparedness and response, verification of arms control agreements, and addressing the needs of communities affected by conflicts.

23. As the Holy See already proposed at the Sixth Review Conference in December 2021, it could also be opportune “to consider establishing an International Organization for Artificial Intelligence, to facilitate […] the fullest possible exchange of scientific and technological information for peaceful uses and for the promotion of the common good and integral human development.”[22] In this regard, it could also be useful to explore the possibility of establishing a body to monitor the peaceful use of artificial intelligence, in analogy with what has been done for Outer Space activities.

24. Given the magnitude of the challenges before us it is essential to show political will before it is too late. Serving peace means actively promoting international agreements capable of safeguarding human dignity and responsibility where life and coexistence amongst peoples are at stake.



[1] Pope Leo XIV, Magnifica Humanitas Encyclical Letter on Safeguarding the Human Person in the Time of Artificial Intelligence (MH), n. 198.

[2]  Cf. Dicastery for the Doctrine of the Faith and Dicastery for Culture and Education, Note of the Holy See on the Relationship Between Artificial Intelligence and Human Intelligence “Antiqua et Nova”, 28 January 2025, n. 99.

[3] Document CCW/GGE.1/2026/3.

[4] Pope Leo XIV, MH, n. 200.

[5] Declaration of the Dicastery for the Doctrine of the Faith “Dignitas Infinita” on Human Dignity, 2024, n. 1.

[6] Cf. Pope Leo XIV, MH, n. 198.

[7] Cf. International Theological Commission, Quo Vadis, Humanitas? Thinking Through Christian Anthropology in the Face of Certain Scenarios for the Future of Humanity, n.2.

[8] Antiqua et Nova, cit., n. 100.

[9] Pope Leo XIV, MH, n. 198.

[10] Cf. Pope Leo XIV, MH, n.198.

[11] Cf. Document CCW/CONF.VI/WP.3, “Translating Ethical Concerns into a Normative and Operational Framework for Lethal Autonomous Weapons Systems”, submitted by the Holy See to the Sixth Review Conference of the CCW, 13-17 December 2021.

[12] Document CCW/GGE.1/2026/CRP.1/Rev.1, par 35.

[13] Cf. Ethical and Moral Dilemmas in the Application of Artificial Intelligence and Lethal Autonomous Weapons Systems, Apostolat Militaire International, September 2026.

[14] Pope Leo XIV, Message for the LIX World Day of Peace, 1st January 2026.

[15] Document CCW/GGE.1/2026/CRP.1/Rev.1, par 28.

[16] Cf. Pope Leo XIV, MH, n. 199-200.

[17] Cf. Pope Leo XIV, MH, n. 199.

[18] Cf. Pope Leo XIV, MH, n. 105.

[19] Pope Leo XIV, Address to the participants in “The Future of AI: A Dialogue between the Vatican and the Scientific World on the Developments and Applications of Intelligent Technologies”, 30 October 2026.

[20] Cf. Doc. CCW/CONF.VI/WP.3 submitted by the Holy See to the Sixth Review Conference of the CCW, cit., n.23.

[21] Pope Leo XIV, Address at the Presentation and Promulgation of the Encyclical Letter Magnifica Humanitas, Synod Hall, 25 May 2026.

[22] Cf. Doc. CCW/CONF.VI/WP.3 submitted by the Holy See to the Sixth Review Conference of the CCW, cit., n.24.