Pope Leo XIV used his first encyclical — a major papal teaching document reserved for matters of broad moral and social importance — to deliver a sweeping warning about artificial intelligence, arguing that AI must serve humanity rather than weaken human dignity, displace workers, or move beyond meaningful human control.
The document, Magnifica Humanitas, was released in the final week of May and it was introduced in an unusually high-profile Vatican presentation that also included Chris Olah, the Anthropic co-founder whose work centers on interpretability – the effort to understand what AI systems are actually doing inside, rather than just observing what they output.
Pairing the encyclical and Olah drew immediate attention. Magnifica Humanitas warned about labor disruption, opaque systems, concentration of technological power, and weapons that move beyond human control. Vatican coverage framed the document as a defense of human dignity, truth, the dignity of work, social justice, and peace. Leo was treating AI as a social and moral question with consequences for workers, institutions, and human responsibility.
What the encyclical argues
The labor theme was especially important. The Vatican tied Magnifica Humanitas to the legacy of Rerum Novarum, Pope Leo XIII’s 1891 encyclical on workers’ rights and industrial society. That gave the new document a clear frame: Just as the Church once addressed the upheaval of the Industrial Revolution, Leo XIV was signaling that AI belongs in that same category of world-shaping change, and he called on AI companies to cool the competition around the technology, especially where it touches labor and warfare.
In a nutshell, the pope warned that AI could deepen inequality, weaken labor, centralize power, and distance human beings from moral responsibility, especially in areas such as warfare and social decision-making.
The emphasis was less on what the systems can do than on who benefits, who loses, and who remains accountable. Two lines from the text help explain its posture. Leo wrote that in this moment humanity must “remain profoundly human,” and he argued that “technology is never neutral,” because it reflects the priorities of the people and institutions that build, finance, regulate, and use it.
The power question runs through the document as well. The encyclical argues that modern technological development is increasingly shaped by private actors whose reach and resources can rival or exceed those of many governments. That helps explain why Leo treated AI as a governance issue, not just an engineering one. In his telling, the risks are not limited to faulty outputs or runaway models. They also include the concentration of decision-making power in systems that can reshape work, public life, and war from behind a wall most people can't see past.
Anthropic's co-founder makes the case
Olah's role added another layer. He was the sole tech-industry representative to appear alongside the pope and speak at the Vatican presentation, where he argued that AI must be guided from outside Big Tech — that governments, religious leaders, and civil society should help steer it. The Anthropic co-founder pointed to three concerns: large-scale job disruption, the fair distribution of AI's benefits, and the difficulty of understanding increasingly opaque systems.
The concern about opaque systems is not incidental. It's the problem Olah has built his career on, and it's the one Leo names in the encyclical when he warns about systems that reshape work and war while remaining hard for ordinary people to understand or challenge. The man invited to stand beside the pope and the encyclical were – on this point – worried about the same thing. Olah was not there simply as a guest from Silicon Valley; he was making the case that some of the guardrails around AI must come from outside the companies building it.
All by itself, Olah's argument made the event notable beyond the encyclical itself. Reuters noted that Leo’s personal participation in the presentation broke with normal papal practice. AP likewise highlighted the pope’s decision to present the encyclical publicly and the unusual appearance of Anthropic alongside him. The event put one of AI’s central tensions in plain view: The Vatican was calling for stronger outside guidance around a technology being built by powerful private companies, while one of those companies was standing beside it making the case.
Critics see "ethics washing"
Reaction to the encyclical and to Anthropic’s presence was mixed. Most coverage treated the document as a serious and potentially influential intervention. Overall, it's fair to say it resonated with many readers because it addressed AI in moral, social, and human terms rather than as a purely technical issue.
But the event itself also drew sharp criticism, much of it aimed less at the encyclical than at Anthropic's place in it. Critics across the AI-policy spectrum accused the company of a kind of "ethics washing" – also dubbed "pope washing" – arguing that a leading AI firm stood to gain reputationally from appearing beside a moral critique of the industry it helps power. The tension was hard to miss: the encyclical's concern for labor and the ecological costs of technology sat uneasily against Anthropic's place inside an energy-hungry, capital-intensive AI boom.
The most pointed objection came from Timnit Gebru, the prominent AI ethics researcher. Writing on LinkedIn, she attacked the Vatican for sharing a stage with a company "about to make billions" from its IPO, and likened the partnership to teaming up with the Sackler family to discuss the harms of OxyContin. Her deeper point was about who was left out: Rather than platform a powerful lab, she argued, the Vatican could have heard from the exploited data workers and the communities fighting AI's data centers — the people most exposed to the technology's harms.
Gebru's criticism doesn't erase the encyclical's significance, but it sharpens the question the event raised. If AI is to be guided by voices outside Silicon Valley, who gets invited into that conversation, and on what terms?
