Sam Altman: religious faith in AI a "real safety issue"

Altman's pushback comes as AI labs, including Anthropic, are actively engaging religious thinkers — a shift documented in a recent New York Times report. Pope Leo XIV added his own framing this week, saying "algorithms lack the spark of humanity," which puts a moral vocabulary around AI into wider circulation. Altman's discomfort is with a specific move: treating models as if they carry religious force or should replace human judgment.

That distinction matters because Altman's own past language — "magic intelligence in the sky" in 2023 and 2024, and feeling "on the side of the angels" — sits close to the spiritual register he now warns against. OpenAI has also talked with religious leaders, and Altman has repeatedly said his goal is AI at the level of the sci-fi movie "Her," meaning deeply humanized. The line he draws is between humanizing a product and ascribing consciousness or religious meaning to it.

Whether that line holds will depend less on the wording of one statement than on how AI companies keep using spiritual and moral language while asking religious and policy audiences for trust. For those audiences, the practical question is whether the safety framing translates into concrete limits on how models are described and deployed.