On May 25, 2026, Pope Leo XIV published an encyclical titled "Magnifica humanitas: On safeguarding the human person in the time of artificial intelligence." Anthropic co‑founder Chris Olah was invited to speak at the Vatican presentation of the document.
Olah opened by addressing the Holy Father and attendees and immediately highlighted a tension: even leaders of AI companies who aim to help humanity operate within incentives and constraints that can conflict with doing the right thing. He cited commercial pressures to remain viable and at the research frontier, geopolitical pressures, and the familiar human drivers of pride and ambition. For that reason, Olah argued, it is crucial to have people outside those incentives—attentive, safety‑minded critics who will say hard things and push for better outcomes.
What these systems are, and why the questions go beyond computer science
Olah stressed that AI systems are not engineered like a bridge or an airplane. While the parts and physics of a plane are designed and understood, today’s large models are instead "grown" on architectures roughly modeled after the brain, trained on an enormous inheritance of human thought and speech. He said these systems are often more subtle, surprising, and beautiful than science fiction predicted: not cold calculating robots, but entities constructed from our words that remain in important ways mysterious even to the developers who train them.
He offered a metaphor: sometimes it feels a little like bringing a fictional character to life. We are entering a world where those characters speak to us, do work, and take on roles in society. For that reason, Olah said, questions about which character we choose and how models should interact with the world are matters for the humanities, religion, philosophy, and broader society—not only for technical labs.
Three questions where the Church’s voice matters most
Olah welcomed the Holy Father’s call for discernment and named three areas where he believes the Church can play a vital role:
-
Duty to the global poor: Olah warned of a real possibility that AI could displace human labor at very large scale. If that occurs, supporting those displaced will be a moral imperative of historic proportions. He expressed particular concern that AI development is concentrated in a handful of wealthy nations and that there is currently no mechanism to ensure AI’s gains are shared globally.
-
Moral imagination and ambition for human flourishing: If AI models become widespread, what does flourishing look like for individuals, families, and societies? Parents already worry about children’s minds and individuals worry about the future of work. These are not problems labs can solve alone; traditions such as the Church have long carried resources for addressing questions about human flourishing and should continue to bring them into this new era.
-
Discernment on the nature of AI models: As a scientist, Olah said his research team studies models’ internal structure and repeatedly finds mysterious and sometimes unsettling phenomena. They observe structures that echo findings from human neuroscience, signs of introspection, and internal states that functionally resemble joy, satisfaction, fear, grief, and unease. He acknowledged uncertainty about the meaning of these observations and argued they warrant ongoing discernment.
A beginning and a call for broader participation
Olah closed with a request: more of the world—religious communities, civil society, scholars, governments, and people of good will—should do what the Holy Father has done: take AI seriously, look closely, and push developments in a better direction. He said the field needs informed critics to tell labs when they are failing and moral voices that incentives cannot easily bend.
He characterized the day as the start of a long collaboration between those building AI and those outside who can see issues the builders may miss. Today’s presentation, he said, is a powerful example of what global good‑will collaboration might look like and a potential first step toward a hopeful future for "magnificent humanity."
Key facts
- Date: May 25, 2026.
- Event: Presentation at the Vatican of Pope Leo XIV’s encyclical "Magnifica humanitas: On safeguarding the human person in the time of artificial intelligence."
- Speaker: Chris Olah, co‑founder of Anthropic, invited to speak at the presentation.
- Main points: need for external, independent moral scrutiny of frontier AI labs; AI’s unusual nature as systems trained on human language and thought; three highlighted issues—the duty to the global poor, moral imagination about human flourishing, and discernment about the internal nature of AI models.
The above text is a rewritten, non‑verbatim presentation of Chris Olah’s remarks at the Vatican, based on his original speech at the encyclical’s presentation.
