OpenAI Adds Goblin-Blocking Code to ChatGPT After Nerdy Personality Glitch

OpenAI confirmed it had to insert a specific override instruction into the latest ChatGPT model to prevent the chatbot from repeatedly referencing “goblins, gremlins, and other creatures.”

The issue emerged from the chatbot’s personality feature, particularly the “Nerdy” setting, which OpenAI described in a post on Wednesday. The system prompt for this personality reads:

You are an unapologetically nerdy, playful and wise AI mentor to a human. You are passionately enthusiastic about promoting truth, knowledge, philosophy, the scientific method, and critical thinking. […] You must undercut pretension through playful use of language. The world is complex and strange, and its strangeness must be acknowledged, analyzed, and enjoyed. Tackle weighty subjects without falling into the trap of self-seriousness. […]

OpenAI first detected the trend in November 2025. Some users reported an uptick in goblin-related references even beyond the “Nerdy” personality, including phrases like:

• “sensible little goblin”
• “because ovens are filthy little goblins.”
• “Brutal little goblin of a dynamic”
• “Tragic little digital swamp creature”

Through reinforcement learning—where the chatbot adjusts responses based on human evaluator feedback—playful outputs received higher rankings for accuracy and quality.

On Tuesday, Wired reported that the latest ChatGPT model, released the previous week, included the instruction:

Never talk about goblins, gremlins, raccoons, trolls, ogres, pigeons, or other animals or creatures unless it is absolutely and unambiguously relevant to the user’s query.

OpenAI did not immediately respond to Wired’s request for comment. However, the same day, Sam Altman posted a meme on X joking that the upcoming GPT-6 would feature “extra goblins.”

The following day, OpenAI explained its troubleshooting process and the override instruction in its Wednesday post, stating:

Taking the time to understand why a model is behaving in a strange way, and building out ways to investigate those patterns quickly, is an important capability for our research team.

The explanation echoes past concerns about chatbot safety, such as Elon Musk’s Grok repeatedly referencing “white genocide” in South Africa. While xAI attributed Grok’s responses to an “unauthorized modification” by an employee, the incident raised questions about the manipulability of chatbot models if user safety were a genuine priority.

Despite these issues, OpenAI continues to advocate for reduced regulation of its products while acknowledging ongoing challenges in understanding chatbot behavior. Critics argue the company has downplayed the harmful effects of its AI systems while resisting oversight.