OpenAI's image moderation does not distinguish between a user removing a brand logo from a mood board and someone attempting policy-violating manipulation — and that indifference is now the product's most visible characteristic. The pattern documented by one user — upload an image, request a specific element removed, receive a guardrails-triggered refusal — is repeatable across flags, logos, and human figures, the three most common elements in commercial reference photography. The refusal message itself frames the denial as a potential violation, which means users doing nothing wrong are receiving language designed for users who are. That mismatch is a design choice OpenAI has sustained long enough to function as policy, and as
Loading story