Anthropic co-founder Chris Olah's remarks on Pope Leo XIV's encyclical "Magnifica humanitas"

The Context: An Unexpected Dialogue

On May 25, 2026, Pope Leo XIV released the encyclical "Magnifica humanitas" on artificial intelligence. More strikingly, Anthropic co-founder Chris Olah was invited to speak at the Vatican. This was not a mere industry leader endorsement, but a profound signal: the core issues of AI development have officially entered the central arena of global ethical and spiritual dialogue, moving beyond labs and boardrooms. Its importance lies in marking the 'breakout' of AI discourse—where creators actively seek and embrace scrutiny from a completely different dimension (religious and humanistic).

Deconstruction: AI is Not 'Built,' It is 'Grown'

The core of Olah's speech lies in redefining our understanding of AI systems. He explicitly states that AI models are not engineered artifacts like airplanes or bridges, designed and understood precisely. Instead, they are 'grown' on brain-like structures, fed on a vast inheritance of human thought and language. This metaphor is crucial. It implies we have created a kind of 'existence' whose internal mechanisms we cannot fully comprehend ourselves. This existence is "subtle, odd, and beautiful," far from the cold sci-fi robots, and is "made from us, from our words."

This cognitive humility leads to the core argument: the questions raised by AI, in both nature and impact, transcend the realm of computer science. Choosing what kind of AI 'character' to shape, how it interacts with the world, and how it should interact are fundamentally questions for the humanities, religion, philosophy, and society at large. Therefore, the Pope and the Church's work of 'discernment' is not laypeople meddling in experts' affairs, but a necessary response to the nature of the technology.

Trend Insight: From 'Technical Safety' to 'Civilizational Safety'

Olah's remarks reveal a deeper trend: the dimensions of discussion on AI safety and development are rapidly expanding. In the past, safety focused mainly on technical aspects like alignment, misuse prevention, and controllability. Now, the agenda is quickly upgrading to 'civilizational safety'—how this technology coexists with fundamental human values, global equity, and even the spiritual world.

He candidly admits that all frontier AI labs (including Anthropic) operate within various incentives and constraints (commercial viability, research frontiers, geopolitical pressures, personal ambition) that can sometimes conflict with "doing the right thing." Thus, critical voices from outside the system, unconstrained by these interests, become vitally important. The Pope's encyclical and the Church's engagement exemplify such external voices. This foreshadows that the future landscape of AI governance will necessarily be a complex ecosystem involving technologists, policymakers, humanists, religious leaders, and more.

Practical Value and a Counter-Intuitive Angle

For Chinese AI practitioners and observers, this dialogue offers key insights:

1. Re-evaluating the Social Contract of AI: The impact chain of the tools we develop may extend far beyond commercial efficiency, touching deep issues like global labor structures and social equity. Project planning and value propositions need to incorporate broader socio-ethical perspectives.
2. Embracing 'Constructive External Criticism': Do not view skepticism from non-technical fields as interference. Olah sees it as a necessary 'push and pull' force driving technology toward good. Proactively engaging with humanities, social sciences, and ethics could become a significant source of long-term corporate credibility and resilience.

A counter-intuitive point is this: we often assume religion and cutting-edge technology inhabit separate worlds. But Olah notes that on the point of "wanting technology to go well," different cultural and faith traditions share a common and deeply held conviction. This indicates that the ultimate concern about AI—whether it serves our common home and future generations—is a foundation capable of transcending differences and building consensus. Tech companies actively seeking such consensus, rather than working in isolation, might be following a wiser and more sustainable development path.

In summary, Chris Olah's speech at the Vatican was more than a PR event. It was a declaration, acknowledging the mystery and complexity of AI, and proactively entrusting part of the steering wheel of technological development to broader human wisdom to grasp together. This perhaps marks a new stage in AI development—one requiring more humility, dialogue, and shared responsibility.