Sam Altman Warns Against Framing AI in Religious Terms

What happened

OpenAI CEO Sam Altman has publicly cautioned against framing artificial intelligence in religious or mystical terms, describing the tendency to attribute spiritual significance to AI models as a genuine safety concern. His remarks come amid wider industry discussion of AI's spiritual dimension, following reports that rival lab Anthropic has held meetings with religious thinkers to explore AI's societal implications, and a recent statement from Pope Leo XIV addressing artificial intelligence.

The comments mark a notable shift in tone from Altman, who as recently as 2024 used language evoking "magic intelligence in the sky" and spoke of feeling aligned with "the side of the angels" in his work on AI development, according to reporting by The Decoder. His current position distances OpenAI from that earlier framing, instead positioning quasi-religious attribution to AI systems as a risk to be managed rather than a narrative to encourage.

Why it matters

How AI leaders talk about their own technology shapes public perception, trust and behaviour toward these systems — and that framing carries real consequences. Language that casts AI as quasi-divine or all-knowing can encourage users to over-trust outputs, defer judgment inappropriately, or form parasocial and even devotional attachments to chatbots and models, rather than treating them as capable but fallible tools.

For organisations deploying AI in customer-facing or decision-support roles, this is a direct reminder that the metaphors used to introduce AI to users and employees are not neutral. They calibrate expectations, trust thresholds and the degree of scrutiny people apply to AI-generated outputs — all of which has operational and reputational stakes well beyond messaging.

The Renascence take

This is a story about narrative discipline as much as it is about AI safety, and the shift is instructive for anyone shaping how AI is introduced to customers or staff.

The interesting signal here isn't that Altman changed his mind — it's that a major AI lab now treats grandiose, quasi-religious framing of its own product as a liability rather than a marketing asset. That's a behavioral-economics lesson hiding in plain sight: the words leaders use to describe AI set the trust calibration users bring to it. Organisations rolling out AI tools to customers or employees should audit their own language with the same discipline — avoid framing that implies infallibility, omniscience or emotional dependency, and instead design onboarding and UX copy that sets accurate expectations about what the system can and cannot do. Trust built on inflated metaphors is brittle; trust built on calibrated honesty about capability and limits is what actually scales adoption responsibly.