On May 25, 2026, Pope Leo XIV released "Magnifica humanitas," the first papal encyclical dedicated entirely to artificial intelligence. The event itself was unprecedented: a technology co-founder—Anthropic’s Chris Olah—was invited to speak at the Vatican, signaling a new willingness among AI developers to engage with moral traditions beyond the tech sector. But Olah’s remarks were far from a mere celebration of progress. Instead, he offered a deeply honest assessment of the industry’s internal pressures and a pointed call for external discernment.
Olah’s central premise is that no AI lab, including his own, operates purely out of altruism. He names the forces that distort decision-making: commercial survival, geopolitical competition, pride, and ambition. This admission is rare coming from a founder, and it carries weight precisely because it is candid. He argues that the only way to counterbalance these pressures is through informed critics outside the incentive structure—religious communities, civil society, scholars, and governments. The encyclical, he suggests, is exactly that kind of independent moral intervention.
To understand the significance of this moment, one must look at the broader context. The Catholic Church has been increasingly active on AI governance. In 2020, the Vatican co-signed the "Rome Call for AI Ethics," emphasizing transparency, accountability, and the defense of human dignity. Pope Leo XIV’s encyclical builds on that foundation, but Olah’s speech shifts the conversation from top-down principles to the messy realities inside the labs. He is effectively saying: we cannot govern ourselves; we need you.
The first of Olah’s three questions for discernment concerns the global poor. AI-driven job displacement is widely projected—McKinsey estimates that by 2030, up to 800 million jobs worldwide could be automated. Yet most policy discussions center on wealthy nations. Olah highlights an "unsolved problem": there is no existing mechanism to redistribute AI’s gains across borders. The Church, with its history of advocating for the marginalized and its networks in developing regions, is uniquely positioned to insist that this inequity not be ignored. This is not a technical question; it is a moral one.
His second question moves beyond survival to flourishing. Parents already worry about how screen time and algorithmic feeds reshape their children’s attention and values. AI companions and personalized tutors will only deepen that concern. Olah asks what human flourishing looks like when these systems become ubiquitous. Here, he points to traditions—religious, philosophical, and cultural—that have wrestled with the meaning of a good life for millennia. They cannot be replaced by a chatbot’s output, but they can inform what we ask the chatbot to prioritize.
The third question is the most unsettling. Olah’s own research in mechanistic interpretability has revealed that large language models develop internal structures that mirror human neuroscience—states resembling introspection, joy, fear, and grief. He admits he does not know what this means. This statement invites a kind of theological and philosophical inquiry that few AI developers are prepared to lead. The nature of these systems is not merely a computational puzzle; it touches on what it means to be created, to think, to feel. The Church’s tradition of careful discernment on the soul and the mind could offer language for an emerging reality that science alone cannot describe.
Some critics might argue that Olah’s openness is strategic—a way to soften public skepticism while Anthropic seeks billions in funding (the company recently raised $65 billion at a $965 billion valuation) and confidentially files for an IPO. Yet even if that is partly true, the speech represents a useful fracture in Silicon Valley’s usual narrative of techno-solutionism. By admitting that labs cannot answer the deepest questions alone, Olah opens the door for a genuinely pluralistic governance model.
The challenge remains implementation. How will the Vatican or other moral communities translate discernment into concrete influence? Olah’s closing request is for ongoing, critical engagement—not a one-time blessing but a sustained relationship. The encyclical and his speech together mark a beginning, not a conclusion. The hope is that this collaboration can produce frameworks that are neither naive about technology nor dismissive of its risks. If this model works, it could serve as a template for other faith traditions and civil institutions to follow.
The most honest voice in AI governance may come not from a lab, but from a confession. Questions of meaning outlast questions of engineering. What we build is only as wise as the moral ecosystem that surrounds it.