Goblins In Your Chat? Why ChatGPT-5.5 Won’t Stop Talking About Fantasy Creatures

OpenAI’s

latest ChatGPT-5.5 is once again making headlines. But this time, the reason is about its capabilities to produce realistic images or generate good graphics in Bengali. Instead, the AI tool is gaining popularity as it often mentions whimsical creatures like gremlins, and goblins in normal conversations. Several social media users have shared screenshots of their chats where ChatGPT mentioned such fantasy creatures.

OpenAI Acknowledges The Glitch

The San Francisco-based company has acknowledged the glitch and it explained that it originated from previous training choices. This model was fine-tuned with various personality styles during the development which also included a ‘nerdy’ tone. This has led to the chatbot generating creative and metaphorical responses. This style often uses fictional elements like goblins and trolls to explain its ideas to users. Over the course of time, the AI began favouring use of such language more as it was regarded during the training. OpenAI explained, “Starting with GPT‑5.1, our models began developing a strange habit: they increasingly mentioned goblins, gremlins, and other creatures in their metaphors. Unlike model bugs that show up through a tanking eval or a spiking training metric and point back to a specific change, this one crept in subtly.”

It explained that a single “little goblin” in an answer could be harmless as well as even charming. Across model generations, though, the habit became hard to miss: the goblins kept multiplying, and the company required to figure out where they came from.

OpenAI Fixes The Glitch

The company further stated, “We retired the “Nerdy” personality in March after launching GPT‑5.4. In training, we removed the goblin-affine reward signal and filtered training data containing creature-words, making goblins less likely to over-appear or show up in inappropriate contexts. Unfortunately, GPT‑5.5 started training before we found the root cause of the goblins. When we began testing GPT‑5.5 in Codex, OpenAI employees immediately noticed the strange affinity for goblins, and we added a developer-prompt instruction⁠(opens in a new window) to mitigate. Codex is, after all, quite nerdy.”As of now, what eventually emerged as an error has turned into a viral moment. As OpenAI continues to fix the issue, netizens continue to enjoy the AI tool’s unexpected responses and share them online.