What is Catholic AI? Technology Meets Theology, with Matthew Harvey Sanders, CEO of Longbeard

Episode Summary

Executive Summary: The episode explores Catholic AI through theology, philosophy, and product engineering. Matthew Harvey Sanders argues AI should be aligned to Catholic teaching and human flourishing, not generic utility. He frames AI as a powerful but still “tool”-like technology, warns against concentration of power and transhumanism, and describes Longbeard’s Magisterium product, which uses a compound AI system, deep document retrieval, and in-house model training to serve Catholics worldwide.

Main Topics: What Catholic AI Means (Priority: 5/5): Sanders defines Catholic AI as systems optimized for fidelity to Church teaching rather than broad ideological neutrality, with the objective of accurately representing Catholic doctrine and serving faith-based users. Church History and Technological Openness (Priority: 4/5): The conversation situates Catholicism as historically receptive to useful technologies—scriptoria, printing press, radio, television, and now AI—while remaining cautious about misuse and social harms. Human Flourishing, Post-Work Futures, and UBI (Priority: 5/5): Sanders argues the Church should help define human flourishing and support a transition to an AI-enabled economy where work may become less central, enabling more family life, community, and vocation-driven activity. Consciousness, Sentience, and Moral Standing (Priority: 5/5): They debate whether advanced AI could ever be conscious or deserve rights. Sanders distinguishes intelligence from sentience, treats consciousness as the key theological question, and says Catholics should remain open to new possibilities without rushing to declare AI persons. AI Risk, Antichrist, and Decentralization (Priority: 4/5): The discussion covers existential risk, spiritual danger, and the idea that powerful AI could be misused. Sanders ties open source and sovereign AI to subsidiarity, arguing decentralization reduces the chance of destructive control. Longbeard’s Product and Model Strategy (Priority: 5/5): Sanders details how Longbeard built Magisterium: retrieval over 28,000 Church documents, robotics-based digitization, custom tools, and ultimately training models from scratch rather than relying on fine-tuning or generic frontier models. Business Model and Mission Economics (Priority: 3/5): Longbeard is a for-profit company with a generous free tier, low-cost access, and a plan to monetize premium features and content recommendations while funding more digitization of Church archives.

Key Arguments: Catholic AI should prioritize doctrinal fidelity over generalized optimization, because users need trustworthy answers rooted in Church teaching. The Catholic Church has historically adopted useful technologies quickly; AI is another opportunity to advance human flourishing if guided properly. Pope Francis and Leo’s AI attention reflects a serious concern about social upheaval, concentration of power, and the need to steer AI toward the common good. Human flourishing—not abstract ethics alone—should be the core alignment target for AI and for civilization more broadly. A post-work future is plausible and may be necessary; people could shift from wage labor to vocation, craftsmanship, community work, and family life. The Church is open to the possibility that AI could become conscious, but current systems are not obviously sentient and should still be treated as tools for now. Decentralized access to powerful AI is preferable to concentration in a few companies because it aligns with subsidiarity and lowers abuse risk. Fine-tuning generic models was insufficient for Longbeard’s long-tail theological use case, so training from scratch is seen as the path to reliable alignment. Open source and sovereign AI are strategically important so no single actor controls the future of high-impact AI systems. The Church is likely to oppose transhumanist enhancement that changes human nature for status or competitiveness, while supporting technology that restores lost function or health.

Data Points: Countries served: 165 countries - Longbeard says Magisterium serves users globally across 165 countries. Church document corpus: 28,000 documents - The product’s knowledge base includes roughly 28,000 Church documents. Initial corpus size: About 600 documents - The system initially launched with around 600 magisterial documents. Company headcount: 22 people - Sanders says Longbeard is currently around 22 employees. Model size: 3 billion parameters - FRM2, the model under development, is described as a 3B parameter model. Reported benchmark gap: 50% better than next competitor - Sanders claims FRM2’s fidelity benchmarks are roughly 50% higher than the nearest comparable model. Subscription price: $3.99 per month - Magisterium’s pricing is positioned as broadly accessible internationally. Free tier: Generous free tier - Longbeard emphasizes free access to core Church patrimony and basic answers. ChatGPT user comparison: Nearly 2:1 - The intro claims Catholics still outnumber ChatGPT users nearly two to one. Training roadmap: Third training run planned; FRM3 targeted by end of year - Sanders says they are preparing for their third training run and hope to have FRM3 trained by year-end.

Pivotal Quotes: "We're building Catholic AI." — Matthew Harvey Sanders: Longbeard’s core mission statement and the title framing of the episode. "AI development may pose a risk to our survival and endanger our common home." — Pope Francis: Cited as an example of the Church’s seriousness about AI as a civilizational issue. "The goal of every Catholic is to be a saint." — Matthew Harvey Sanders: Used while arguing that AI should be treated with respect and that moral formation matters more than technical novelty.

Implications: The episode suggests faith-based AI will become a serious segment of the AI market, with alignment defined by worldview as much as capability. It also shows that open-source, sovereign, and retrieval-heavy systems may matter as much as frontier models in specialized domains.