OpenAI explains why ChatGPT suddenly loved goblins

OpenAI CEO Sam Altman is currently embroiled in a legal drama, but OpenAI Lab engineers just resolved a minor drama (actually a comedy) with ChatGPT.

OpenAI has published a long and detailed report on its website about goblins. No, really. Since the release of GPT 5.1, ChatGPT models have developed a strange fixation on goblins and gremlins, regularly bringing up both in contexts where it doesn’t make sense. Users took notice and OpenAI noticed it too, sparking an investigation. You can read the full report if you want all the juicy details, but basically it all comes down to a misunderstanding of what nerds are.

If you don’t use ChatGPT, the AI ​​chatbot offers a variety of “personalities” you can choose to tailor the tone of its responses. One of these personalities was (until recently deleted) “nerdy.” According to OpenAI’s research, while “cheesy” responses made up only 2.5% of all ChatGPT queries, 66.7% of all mentions of goblins came from that 2.5%. From there, these responses became “rewarded” and spread to other personality types beyond “nerdy.”

“Rewards were only applied in the Nerdy condition, but reinforcement learning does not guarantee that learned behaviors remain perfectly adapted to the condition that produced them,” OpenAI wrote. “Once a style tic is rewarded, subsequent training can diffuse or reinforce it elsewhere, especially if these results are reused in supervised or preference tuning data.”

Anyway, after all that, OpenAI removed the “nerdy” persona in March and instructed its latest models not to mention goblins or gremlins unless it makes sense.