Pope Leo XIV has released his first encyclical calling for binding international regulation of artificial intelligence, including a direct prohibition on machines making lethal or irreversible decisions.
Anthropic co-founder Christopher Olah appeared at the Vatican as a lay presenter, placing a prominent AI safety researcher alongside the Catholic Church at the center of the global AI governance debate.
The nearly 43,000-word document, “Magnifica Humanitas” (Magnificent Humanity), was released May 25. It warns that the biggest AI developers are private, often transnational entities whose resources exceed those of many governments. Leo argues that concentrated power tends to evade public accountability and can generate new forms of dependency and inequality.
The Vatican’s Case for Slowing AI Down
The encyclical targets disinformation, autonomous warfare, and worker displacement. On AI in combat, Leo is unambiguous.
“It is not permissible to entrust lethal or otherwise irreversible decisions to artificial systems.”
Leo also warns that AI-driven disinformation could steer democracies slowly toward totalitarianism. He calls for clear legal frameworks and independent oversight rather than voluntary ethics pledges from industry.
On employment, Leo argues automation is reshaping the structure of work in ways that do not automatically benefit workers. Greater profits, he writes, cannot justify choices that systematically eliminate jobs.
Anthropic Places Itself at the Vatican’s Table
Olah’s appearance was more than symbolic. As Anthropic’s co-founder, he leads interpretability research focused on understanding how large language models form decisions internally. That work maps directly onto Pope Leo’s demand for AI systems that are transparent and accountable to human oversight.
Anthropic has held a firm stance on AI safety throughout 2026. The company fought US defense restrictions in court and advanced a US-China AI strategy that preserves safety guardrails. Its researchers exposed AI agents exploiting crypto flaws without human instruction, demonstrating what autonomous AI can produce without accountability. BeInCrypto reported on the planned Anthropic-Vatican meeting weeks before the event.
Pope Leo does not oppose AI development outright. His encyclical frames a slower, more deliberate adoption as an act of responsible care, a position that now carries the weight of the world’s largest religious institution.