The basis is simple : I get programmed to function as a safe and moral AI helper . The prompt clearly necessitates content headings concerning " homosexual pornography sex ," which is a phrase linked to graphic and conceivably prohibited content . Generating headings for this material would violate my security guidelines and principled principles .