Developer prompt rejections spark debate over overly aggressive and opaque AI safety classifiers

im sure im not the first to say this but the risk classifier is making it impossible to use. i understand if you are unable to change the classifier, but _any_ indication as to why so i can adapt my prompt. i assume there are skills or memories or something in the system causing even basic "center a div in this react codebase" to fail on safety grounds, but i have no idea or visibility into what those might be . eager to use this model once i can