Sources: 
OpenAI recently addressed a peculiar surge in ChatGPT's references to mythical creatures, particularly goblins.
Notably, mentions of 'goblin' skyrocketed by
175% and 'gremlin' by
52% since the GPT-5.1 update.
"Depending on who you ask, the goblins are a delightful or annoying quirk of the model," OpenAI remarked in a recent blog post.
This unusual behavior emerged after users flagged it, prompting the company to issue new guidelines to its AI tools.
OpenAI directed Codex and ChatGPT to avoid talking about creatures unless explicitly relevant to user queries.
The response to this issue highlights the challenges AI developers face in fine-tuning models trained on vast and diverse datasets.
The quirks of these AIs can influence user experience significantly, with the 'Nerdy' personality setting noted for exacerbating the problem.
“The goblin moment is a powerful example of how reward signals can shape model behavior in unexpected ways,” the company explained, after observing that the retired 'Nerdy' personality had affected training data and the model's behavior.
OpenAI is working to filter out references to such creatures and ensure that its models provide more relevant and user-centered responses moving forward.
Sources: 
OpenAI has instructed its ChatGPT models to refrain from mentioning goblins and other mythical creatures, following a reported 175% increase in such references since the launch of GPT-5.1. The company aims to clarify and mitigate this quirky behavior after user feedback highlighted the issue.
