The pope is warning about AI. Anthropic is asking religious thinkers for help

Anthropic, the AI company behind Claude, convened 15 religious thinkers in late March to explore how centuries-old moral traditions could guide its chatbot’s behavior. The discussions followed Anthropic’s realization that its in-house expertise was insufficient to address the ethical complexities of AI, particularly as models like Claude become more persuasive and harder to govern with simple rules. Participants included Greg Cootsona, co-founder of AI and Faith, and Brian Patrick Green, director of technology ethics at Santa Clara University, who noted that religious frameworks offer millennia of experience in shaping moral agents. The meetings aimed to refine Claude’s 'constitution,' a set of principles defining how the AI responds to queries on sensitive topics like end-of-life care or grief. Anthropic’s co-founder Christopher Olah attended a Vatican event in May where Pope Leo XIV unveiled *Magnifica humanitas*, a 42,300-word encyclical warning against AI’s unchecked technical power. The pope framed AI as needing to be 'disarmed'—not rejected but freed from the assumption that its capabilities automatically justify control. The collaboration reflects a broader industry reckoning with AI ethics, where few regulations or standardized methods exist. Religious input could help Anthropic navigate dilemmas like balancing values across diverse societies, though critics question whether theological perspectives align with secular ethical frameworks. The stakes are high: hundreds of millions interact with AI weekly, and the values embedded in these systems influence decisions on critical issues. Anthropic’s outreach signals both humility and improvisation in an industry racing to define ethical boundaries. While religious thinkers emphasize moral formation as a core strength, skepticism remains about whether their insights can translate directly into AI governance. The experiment underscores the tension between rapid technological advancement and the need for foundational ethical guardrails.