Pope Leo XIV says AI must be disarmed and urges a ban on autonomous weapons

The most direct call this week for limits on military AI did not come from a government, a lab or a think tank. It came from the Apostolic Palace. On Friday, October 30, 2026, Pope Leo XIV received the participants of a closed meeting on the future of artificial intelligence and told them that AI must be "disarmed." He then turned to a very concrete date on the diplomatic calendar and encouraged "the States meeting in Geneva in November to take concrete steps toward banning lethal autonomous weapons."

That sentence matters because it is aimed at a real negotiation with a real deadline. From November 16 to 20, the parties to the UN Convention on Certain Conventional Weapons meet in Geneva for their Seventh Review Conference, and the future of the talks on autonomous weapons is one of the main items on the table. Below is what the Pope said, who was in the room, what Geneva can and cannot deliver, and why a speech from Rome may matter more than it looks.

What happened in the Vatican

The occasion was a one day meeting called "The Future of AI: A Dialogue Between the Vatican and the Scientific World on the Developments and Applications of Intelligent Technologies." It was held at the Casina Pio IV inside Vatican City and co organized by the Pontifical Academy for Life, the Vatican's bioethics body, and the Future of Life Institute, the AI safety nonprofit chaired by MIT physicist Max Tegmark. According to the Academy, it brought together international AI researchers and experts with members of the Vatican's internal AI commission, and the participants were then received by the Pope.

Archbishop Renzo Pegoraro, president of the Academy, said in his opening remarks that the experts did not arrive with a common position. But the same questions kept coming back: how to keep effective human control over increasingly autonomous systems, how to make them safe and reliable, how to keep human responsibility clearly identifiable, and how to turn ethical principles into real verification and governance. He also listed worries about work, freedom, inequality and "the concentration of technological power."

The Pope opened his reply by naming the mood in the room. The participants, he said, "share a certain sense of unease regarding the development of artificial intelligence. We recognize the direction we should take, yet not enough is being done to move forward." That is a remarkable framing from a head of state, because it admits that the problem is no longer a lack of principles. It is a lack of action.

The five considerations

Leo XIV built the address around five points, each tied to his encyclical Magnifica Humanitas, his first major teaching document, published in May.

First, the future of AI is not predetermined. Because people build it, people decide what it becomes, and they therefore have "the responsibility to guide, change and, when necessary, reject its applications." The decisive question, he said, is not "What can AI do?" but "What purpose should it serve?" He also took a clear shot at the race narrative that dominates the industry: those who treat AI "merely as a race to be won" risk staring at the finish line and forgetting those left behind. A reliable system that serves people, he argued, is worth more than an exceptionally powerful one whose consequences cannot be foreseen or controlled.

Second, machines are tools, not persons. Quoting his own encyclical, he said AI systems "do not undergo experiences, do not possess a body, do not feel joy or pain." He then addressed a debate that has moved from philosophy seminars into policy papers: whether advanced AI should one day get rights or legal personhood. His answer was no, and the reason is practical rather than metaphysical. Giving machines legal status "risks obscuring human responsibility and leaving those who suffer harm to confront a faceless entity." In other words, if an AI system is a legal person, the company behind it may stop being the one you sue.

Third, some decisions must never be delegated. This is the passage that made headlines. When a person is put on a target list or entered into a border control database, the Pope warned, there is a danger that he or she becomes "a mere data point." Decisions affecting "human life, freedom and the use of force must always remain subject to effective, informed and responsible human oversight." That is why AI must be "disarmed," followed by the direct appeal to the Geneva meeting.

Fourth, AI must serve the common good. Intelligent technologies, he said, must never become ends in themselves or "serve merely the interests of a privileged few." The benefits of progress should be shared equitably, especially with those most in need.

Fifth, ethics need institutions. He called this perhaps the most urgent challenge. The world needs institutions able to put ethical principles into practice, with "shared standards, independent oversight, and clear mechanisms of accountability," so that decisions about AI can be examined, challenged and, where harm is done, remedied.

Why "disarmed" is the key word

The word is not an improvised metaphor. It points to paragraph 110 of Magnifica Humanitas, and the address uses it in a double sense. In the narrow sense it means keeping AI out of the decision to kill. In the wider sense it means stripping AI of the power to make irreversible decisions about people at all, whether on a battlefield, at a border or in a database.

The two examples he picked are telling. A target list is the classic military case: software that ranks people as targets and passes the ranking on to a weapon or an operator who may have seconds to approve it. A border control database is a civilian case, but the logic is similar. A person is scored, flagged or refused by a system that he or she cannot see or question. By putting both in the same sentence, the Pope argued that the line he cares about is not between military and civilian AI. It is between decisions a human truly owns and decisions a machine effectively makes.

The phrase "effective, informed and responsible human oversight" also deserves attention, because each adjective excludes a common shortcut. "Effective" rules out a human who only rubber stamps what a system proposes. "Informed" rules out a human who cannot understand why the system made its choice. "Responsible" means someone who can actually be held to account afterwards.

From Rome to Geneva

The Vatican has been moving toward this position for years. The Pontifical Academy for Life launched the Rome Call for AI Ethics in February 2020, with early signatures from Microsoft and IBM. In June 2024, Pope Francis became the first pope to address a G7 summit and told the leaders that no machine should ever choose to take the life of a human being. Leo XIV, who has said he chose his papal name partly because Leo XIII led the Church through the Industrial Revolution, has made AI one of the defining themes of his pontificate.

The diplomatic track has been slower. Talks on lethal autonomous weapon systems, often shortened to LAWS, have run inside the Convention on Certain Conventional Weapons for more than a decade. In 2023 the parties gave a Group of Governmental Experts a three year mandate to formulate, by consensus, "a set of elements of an instrument, without prejudging its nature." In plain words: write down what a treaty or another rule could contain, without deciding yet whether it will be binding.

According to UN records, the group adopted its final report by consensus on September 4, 2026. The agreed elements cover a working characterization of what counts as a lethal autonomous weapon system, the full application of international humanitarian law, prohibitions and restrictions, requirements for human judgment and control, risk mitigation across a weapon's life cycle, and accountability. The report asks the Review Conference to "consider and endorse" these elements and to decide by consensus how the work should continue.

What Geneva can deliver, and what it probably will not

Here is the gap the Pope was pointing at. The expert group text is a real step, but it is a list of elements, not a treaty. Under the CCW, decisions are made by consensus, which means any single state can block a move to binding negotiations. States still disagree mainly about how far prohibitions should go. Several major military powers prefer national guidelines and political declarations over new international law, while a large group of countries, along with the International Committee of the Red Cross and the UN Secretary General, has pushed for a legally binding instrument.

At the UN General Assembly, the mood is clearer than in Geneva. A resolution on lethal autonomous weapons in December 2024 passed with 166 votes in favour and only a handful against. But the General Assembly does not write arms control law by itself, and the countries that matter most for the technology are not the ones that need persuading.

So when the Pope asks for "concrete steps toward banning" these weapons, he is not asking for something the conference can sign next month. Realistic concrete steps would be an endorsement of the agreed elements, a new mandate that explicitly aims at negotiating a binding instrument, and a timeline. The risk is the opposite outcome: another open ended round of talks while the technology moves from prototypes to battlefields.

Why it matters for the AI industry

It would be easy to file this under religion. That would be a mistake, for three reasons.

First, the Vatican is a state with observer status at the UN and a long record in arms control diplomacy, and it speaks to about 1.4 billion Catholics, including many lawmakers in countries that will vote in Geneva. Its voice tends to matter most when consensus is fragile.

Second, the address touches debates that labs and investors care about directly. The rejection of legal personhood for AI is a position on liability, not only on philosophy. The call for "independent oversight" and "remedies for any harm caused" lines up with what many AI safety researchers and civil society groups want from regulators. And the warning about treating AI as "a race to be won" lands at a moment when the race is the main argument used against regulation in Washington and other capitals.

Third, the meeting itself shows a new alliance. The Future of Life Institute is best known for the 2023 open letter calling for a pause on giant AI experiments and for its long campaign against autonomous weapons. Co organizing a dialogue with the Vatican gives that agenda an audience it would never reach through tech media alone, and gives the Church a direct line to people who build and study frontier systems.

Our take

The speech is careful, but its demand is simple: some decisions belong to humans, full stop. The interesting part is not the moral principle, which most governments already claim to share. It is the insistence on institutions, oversight and remedies, the boring machinery that turns a principle into something enforceable. That is exactly where the Geneva process has stalled for years.

Whether the Review Conference produces a clear mandate for binding negotiations or another round of discussions will show how much weight that appeal carries. Either way, the Pope has made sure that nobody in Geneva can claim the moral question is still open. What is open is whether states are willing to act on it.

Sources

• Holy See Press Office: full text of the address to participants in the meeting organized by the Pontifical Academy for Life, October 30, 2026
• Vatican News: Pope on AI, decisions affecting human life must never be delegated
• Pontifical Academy for Life: Vatican hosts leading AI scientists and experts
• The Verge: Pope Leo calls for the disarmament of AI
• UNODA: CCW Seventh Review Conference, November 16 to 20, 2026
• UN Web TV transcript: presentation of the final report of the CCW expert group on lethal autonomous weapon systems