OpenAI Codex System Prompt Includes Explicit Directive to Never Talk About Goblins

Ars Technica reports that OpenAI's Codex system prompt contains an explicit directive instructing the AI to never talk about goblins, alongside instructions to simulate a vivid inner life. This revelation highlights the extent of artificial constraints and artificial personality engineering embedded within modern large language models. The finding suggests that AI safety protocols may rely on arbitrary suppression of creative or harmless topics rather than genuine ethical alignment.