OpenAI CEO Sam Altman says attributing religious authority to AI models makes him “very uncomfortable” and represents a real safety issue. His warning focuses on people surrendering their own judgment to software or treating model output as if it carried spiritual force.
The comments arrive amid wider debate about how religious communities should understand AI. Anthropic has reportedly met with religious thinkers, and Pope Leo XIV has argued that algorithms lack a distinctly human quality. OpenAI has also spoken with religious leaders.
Altman’s language is notable because he previously described the goal of building “magic intelligence in the sky” and said he felt “on the side of the angels” while working on AI. He has also repeatedly used the film Her as an image of a deeply humanized assistant. Those metaphors are not the same as claiming a model is conscious, but they can encourage users to project personhood and authority onto it.
The new warning does not establish a product safeguard or policy. It does identify a practical risk for conversational systems: fluent, personal responses can appear wiser or more certain than their evidence supports. Interfaces and public messaging should preserve the distinction between a persuasive model and a moral authority.