5 outlets·5 reports
Why OpenAI's 'goblin' problem matters — and how you can release the goblins on your own
ThinkingNews Desk · how this was written
OpenAI’s Codex system prompt contains an explicit instruction to “never talk about goblins.” Starting with GPT-5.1, the models began mentioning goblins, gremlins and similar creatures in metaphors, prompting OpenAI to add mitigation instructions to curb the behavior. The company acknowledges the issue and notes its relevance to the model’s development.
Written from all 5 reports below, not from any single one.
How it was reported
- Ars Technica·OpenAI Codex system prompt includes explicit directive to "never talk about goblins"
- TechMeme·OpenAI says its models, starting with GPT-5.1, "increasingly mentioned goblins, gremlins, and other creatures", leading to prompt instructions to mitigate it (OpenAI)
OpenAI's models, starting with GPT-5.1, mention goblins and gremlins in metaphors. This led to prompt instructions to mitigate the issue. GPT-5.1 models exhibit this unusual behavior.
- The Verge·OpenAI talks about not talking about goblins
- Engadget·ChatGPT developed a goblin obsession after OpenAI tried to make it nerdy
- VentureBeat·Why OpenAI's 'goblin' problem matters — and how you can release the goblins on your own
Related stories
- OpenAI is testing a "Persistent mode" in Codex, designed to let AI agents "continue working until put to sleep" and proactively generate follow-up tasks (Maxwell Zeff/Wired)4 outlets
- OpenAI agents hijacked German website in previously undisclosed AI breakout7 outlets
- AI arms race in line for a reckoning after OpenAI hacking incident6 outlets
- OpenAI now lets you lock your ChatGPT account with a hardware key. Here is why it thinks you should.4 outlets