Sam Altman calls assigning religious force to models a safety issue

OpenAI CEO Sam Altman wrote on X on October 3 that he was uncomfortable with attempts to assign religious force to AI models or replace human judgment with them. Axios connected the statement to reports that Anthropic co-founder Christopher Olah had met religious and philosophical leaders while exploring model consciousness and moral behavior.

The debate matters to product teams because the language used for models affects trust, consent and responsibility. Treating a model as a moral authority can hide the human choices behind training, deployment and policy, while dismissing user beliefs can also damage adoption.

Keep claims about agency and consciousness separate from system behavior. In product copy and evaluation, state what the model can observe, what it can decide and who remains accountable for the result.