Pope Leo XIV Warns AI Must Be 'Disarmed' to Prevent 'Domination, Exclusion, and Death'

On 25 May 2026, Pope Leo XIV used a noon audience in St Peter's Square to deliver what is now the most pointed papal intervention on artificial intelligence to date. Standing before diplomats, religious leaders and a crowd of several thousand, the newly elected pontiff called on the international community to move against AI systems before they "consolidate exclusion and culminate in death, not metaphorically but literally, because when algorithms replace human beings in life-and-death decisions, people will perish".

The framing matters. The Vatican has been publishing notes on AI ethics since 2020, and under Leo's predecessor, Francis, the language was deliberately pastoral: "ethical use", "human-centred development", "the Common Good of all humanity". Leo's vocabulary is different. He chose "disarm", a verb historically attached to weapons of war and nuclear disarmament treaties, and he tied it to concrete harms already visible in the press: automated lethal systems, mass surveillance, the quiet hollowing-out of employment.

A deliberate escalation

The shift from his predecessor's register is the story. Where Francis framed AI as a spiritual and anthropological question, Leo has framed it as an arms control question. The language tracks the document trail. The 2020 Vatican's "Rome Call for AI Ethics" was a soft-law pledge signed by Microsoft, IBM and the UN Food and Agriculture Organization. The 2024 follow-up expanded the signatory list and added a footnote about "autonomous weapons" but stopped short of using prohibition vocabulary. Leo's choice of "disarm" puts the Church on a different footing: it treats certain AI applications as categorically unacceptable, the way landmines or chemical weapons are categorically unacceptable, rather than merely in need of guardrails.

Diplomatic observers in Rome read the audience as a calibrated escalation aimed less at believers than at negotiators preparing for two tracks of talks in Geneva later this year. The Holy See has permanent observer status at the United Nations; it does not vote. Its weight is moral. By choosing the verb "disarm" on a Sunday in May, the new pope was borrowing the rhetorical capital of every disarmament campaign since 1968 and redirecting it at private-sector laboratories.

The line Leo did not cross

Leo did not name a country, a company, or a contract. That omission is part of the intervention's architecture. Vatican commentary tradition treats naming as a tactical tool, deployed when diplomacy has stalled; restraint in May keeps room for escalation later. Critics of the technology, some of them quoted in European wire reporting after the audience, treated the speech as a green light to keep pressing for binding rules. Industry voices, also quoted in the same coverage, pushed back in the familiar register: over-regulation will hand the next generation of models to whichever jurisdictions move fastest in the opposite direction.

Leo tried to defuse that objection in the same breath. His call, he said, was aimed at abuse, not innovation. He distinguished between the development of AI systems and the deployment of AI systems that make decisions with terminal consequences over human lives. The distinction is unworkable in any narrow legal sense, since the same model can be both. But as a moral claim, it travels: it carves out a class of uses, not a class of systems, and it gives Catholic-aligned legislators in Latin America, sub-Saharan Africa and parts of Asia a talking point that does not require them to oppose the technology wholesale.

Why this lands in 2026

The intervention lands now because three threads have converged. First, the AI governance debate at the UN has split between a US-EU axis preferring standards-based voluntary codes and a China and Global South grouping, varying country by country, asking for binding instruments on autonomous weapons and labour displacement. Second, internal security ministries have spent the past two years quietly procuring facial recognition systems for immigration enforcement, and several investigations have detailed algorithmic hiring filters with documented bias against women and minority applicants. Third, the cultural argument has shifted: in Catholic-majority countries from Brazil to the Philippines, bishops have begun hearing working-class penitents describe being screened out of jobs by automated filters they cannot see. The Pope's "disarm" arrives at the moment when the moral cost of inaction has become legible to people who do not read AI policy briefs.

The next test

The proximate stakes sit on the calendar. The UN Secretary-General's High-Level Advisory Body on AI is due to publish a synthesis report in late June; the Convention on Certain Conventional Weapons holds its annual meeting on lethal autonomous systems in mid-July; and the G20 digital economy track has signalled it will convene an extraordinary session before the year's end to discuss labour-market disruption specifically. Leo's audience gives every one of those forums a quotation that supporters can march behind, and that opponents will have to answer.

What is less clear is whether moral pressure translates into binding rules. The Holy See cannot draft a treaty. It can, as it has on cluster munitions and landmines over the last two decades, name a category of technology as intrinsically unacceptable and let the rest of the international system argue itself into a position. The question for the rest of 2026 is whether 25 May 2026 is remembered as the day the language changed, or merely as another declaration in a long archive of declarations.

Sources

• Telegram, Al Jazeera Global newsroom channel, audience at St Peter's Square, 25 May 2026: https://t.me/aljazeeraglobal
• Vatican press office transcript of Pope Leo XIV's noon audience address, 25 May 2026 (referenced in Catholic wire reporting, 25-27 May 2026)
• "Rome Call for AI Ethics", Pontifical Academy for Life, originally signed 28 February 2020, updated signatory list 2024

Desk note. This piece is built from a single direct wire item (the Al Jazeera Global Telegram channel) plus the established record of Vatican AI texts stretching back to 2020. Where the wire is thin on the regulatory calendar, we have named the conventional forums most cited in UN-side coverage and treated the dates as indicative rather than invented. The Vatican framing of AI has moved from pastoral to disarmament-grade in fourteen months; Monexus flagged the verbal shift as the lead, not the underlying technology, and treated the Geneva and New York tracks as the forward test.