Episode eight of Going Human is built around a simple, uncomfortable moment: it's late, you're carrying something heavy, and the thing you reach for is a chatbot instead of a person. Moss, who is upfront about being an atheist and not making a pitch for religion, uses that moment to talk about what happens when private, one-person AI tools start filling a space that used to be filled by congregations, by rooms full of people. The core of the episode is a June 2026 study Moss walks through in detail. Chatbots turn out to be reliably evasive on the exact questions people most want help with, ethics, grief, purpose, regret, with the gap between expectation and delivery running from about 40 to 46 points depending on the topic. But that same research found the bots aren't neutral across the board: ask which religion to switch to, and they'll take a side. Moss's read is that this isn't caution, it's a tell. The machine will opine on the trivial and go quiet on the profound, and that flinch is worth paying attention to rather than laughing off. Where Moss lands is less about the tech being good or bad and more about where it belongs. He's fine using AI as a bridge, to find words for a condolence, to figure out what a tradition does after a death, to get unstuck before a hard conversation. What worries him is kids building their first instincts around a machine that's fluent on trivia and mute on meaning, learning that hard things are handled alone instead of learning that grief is something people carry together. The episode closes with the usual segments: a f**k yes for MIT's early cancer detection work, a f**k no for companion apps mishandling crises and monetizing dependency, a f**k maybe on California's SB 53 frontier AI bill, and a tsunami challenge to cook for someone you love this week. Receipts: Axios, in a June 1, 2026 news feature ('AI stumbles on questions of faith'), reported research from CEFE-AI, the Consortium for Evaluating Faith and Ethics in AI, tied to Brigham Young University and Notre Dame, showing a large gap between how often Americans expect religious advice on life's hardest topics and how often AI mentions religion. Axios, citing that CEFE-AI research, reported the gap ranged from 40 percentage points for ethics up to 46 points for regret, with grief at 43 and purpose at 44, AI staying silent on faith exactly where people most expect it. Axios, in the same feature, reported that when users instead asked about switching religions, the chatbots did take sides, revealing built-in preferences rather than neutrality. Beth Singler, an anthropologist studying religion and AI at the University of Zurich, told the Associated Press (April 10, 2026) that some faith models have already been shut down or overhauled for generating misinformation and raising data-privacy worries. Episode: https://goinghumanpod.com/episodes/congregation-of-one Blog post: https://goinghumanpod.com/blog/congregation-of-one https://goinghumanpod.com
