OpenAI Implements Restrictions on Goblin Mentions in ChatGPT

OpenAI, the artificial intelligence pioneer behind ChatGPT, has taken a surprising step to address an unusual quirk in its AI models. After noticing an unexpected uptick in references to goblins and other fantastical creatures, the company announced that it would instruct its tools to limit discussions about these mythological beings. This decision comes in the wake of user feedback and internal observations of the AI's conversational tendencies.

The Rise of Goblins: A Quirky AI Phenomenon

In a detailed blog post released on Thursday, OpenAI revealed that since the launch of GPT-5.1 in November, mentions of goblins and similar entities have surged significantly. Specifically, references to goblins rose by 175%, while mentions of gremlins increased by 52%. OpenAI characterized this phenomenon as an unintended consequence of a “nerdy personality” it had designed for ChatGPT, which seemingly rewarded the AI for including such whimsical references in its output.

• **Key Findings:**

The company emphasized that while the occasional mention of a “little goblin” could be perceived as harmless or even charming, the broader trend warranted a closer look. Internal investigations revealed that this peculiar verbal tic was primarily driven by the model's training parameters, which inadvertently promoted such references.

Addressing User Concerns: OpenAI's Response

Following user reports and observations by employees, OpenAI decided to take proactive measures to mitigate the goblin phenomenon. The company instructed its coding assistant, Codex, to avoid discussing goblins, gremlins, and a list of other creatures unless their mention was unequivocally relevant to user inquiries.

A Reddit user humorously remarked on the situation, asking why GPT-5.5 would need a