ChatGPT's Image Tool Is Breaking Its Own Policy Logic

The institutional problem here is not a bug in the conventional sense — it is a moderation architecture that treats provenance as unknowable and defaults to restriction. OpenAI's image generation policies were built to prevent the reproduction of other people's likenesses and copyrighted material. But a filter that cannot verify authorship at all will eventually block the one class of image it was never meant to restrict: a person's own face. The users hitting this wall are not edge cases; they are the baseline user the feature was designed…