At the time the company found that the goblins \u201cdid not look especially alarming,\u201d it said in a statement, but as newer versions of ChatGPT rolled out, they noticed an even bigger uptick.<\/p>\n
For most people, ChatGPT\u2019s abundance of goblins is a benign AI quirk. And the company has since taken measures to crack down on the hordes of goblins in its system, including by issuing a stop gap command that essentially forbids the model from using the word \u201cgoblin\u201d in most conversations.\u00a0<\/p>\n
\tNortheastern Global News, in your inbox.<\/p>\n
Sign up for NGN\u2019s daily newsletter for news, discovery and analysis from around the world.<\/p>\n
![\"\"](\"https:\/\/www.newsbeep.com\/il\/wp-content\/uploads\/2025\/12\/EmailGraphic.png\")
<\/p>\n
But technology experts said the glitch reveals cracks in the foundation of how these systems are trained, and how companies are struggling to keep up with the demands of the AI arms race.<\/p>\n
\u201cIt\u2019s a pressure cooker,\u201d said Christoph Riedl<\/a>, a professor of computer science, information systems and network science at Northeastern University. \u201c[Companies] are under pressure to release new models. They have limited resources and capacity to test things. The processes are super long and complicated. That\u2019s exactly why you see things like this.\u201d<\/p>\n But where exactly did the goblins come from? According to Riedl, the goblin problem stems from how models like ChatGPt get trained.<\/p>\n Specifically, according to Reidl, it\u2019s likely a later training stage called fine tuning, where humans provide feedback to the model on the quality of its responses. That quality is subjective: Users might like a response because of its accuracy, tone or how much it reinforces their beliefs<\/a>.<\/p>\n \u201cThese are reinforcement signals [from the users] to the models to say, \u2018Hey, if I generate an answer that looks like this, then [you] get positive rewards, and if it\u2019s an answer that looks like this other thing, then there\u2019s less of these rewards,\u2019\u201d Riedl said.\u00a0<\/p>\n ChatGPT\u2019s reward system is partially based on its personality customization feature, which lets users choose a tone and style for their version of the chatbot. Personality profiles range from cynical to friendly, but the one that spawned ChatGPT\u2019s goblin problem was the \u201cnerdy\u201d personality, which started using the creatures in metaphors, OpenAI explained in a statement. This version of ChatGPT is designed to be more playful and \u201ctackle weighty subjects without falling into self-seriousness, according to OpenAI\u2019s system prompt for its model.\u00a0<\/p>\n Once a model latches onto a rewarded behavior, it will try to \u201creward hack\u201d as it tries to find shortcuts and generate responses that will get the most rewards. OpenAI might have a broader, richer understanding of what \u201cnerdy\u201d means, but the model \u201cmight optimize for it in a very narrow way that\u2019s not at all what you intended,\u201d Riedl said.<\/p>\n And that\u2019s what seems to have happened. Between December and March, mentions of goblins increased by 3,881.4% in responses from the nerdy personality, according to OpenAI.<\/p>\n But these behaviors can also spread. Once the model starts to see a specific tic being rewarded, even in a specific part of its operations like the nerdy personality, the behavior gets reinforced in later training for the entire model, Riedl said.\u00a0<\/p>\n Perhaps unsurprisingly, OpenAI noted that references to goblins started to pop up in ChatGPT\u2019s other personality profiles.\u00a0<\/p>\n It\u2019s also why, prior to OpenAI rolling out a command that essentially banned ChatGPT from mentioning goblins, the most recent version of ChatGPT continued to mention goblins, alongside other creatures like gremlins, ogres, trolls, raccoons and pigeons. Most references to frogs were still legitimate, OpenAI stated.<\/p>\n Riedl noted that the way this lexical tic spread reveals a worrying trend.\u00a0<\/p>\n He explained that companies will commit an entire datacenter to training their model for months, but have little influence over what happens once the training starts. If goblins or some other unwanted behavior are somehow embedded in that training process, like they seemed to have been here, the company won\u2019t find out until months later.<\/p>\n OpenAI ultimately implemented a quick fix that addressed the issue in the short-term, retiring the \u201cnerdy\u201d personality. But with the demand to create better models more quickly and frequently behaviors like this will continue to slip through the cracks, Riedl said\u00a0<\/p>\n It\u2019s created a situation that \u201cevery [AI] safety researcher is worried about,\u201d Riedl said, one that, at best, produces goblins. Grok, Elon Musk\u2019s AI chatbot, had its own fixation last year: baseless claims of \u201cwhite genocide\u201d<\/a> in South Africa.<\/p>\n \u201cThis time it\u2019s goblins and next time it\u2019s something else that will probably just not go away,\u201d Riedl said. \u201cWe\u2019re lucky if it\u2019s goblins as opposed to white supremacy or [information on] chemical weapons \u2026 or encouraging people to commit suicide<\/a>.\u201d\u00a0<\/p>\n","protected":false},"excerpt":{"rendered":"Starting sometime in November, people who used ChatGPT began noticing some peculiar behavior: the AI chatbot would not…\n","protected":false},"author":2,"featured_media":428773,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[6],"tags":[345,343,4193,201242,85,46,125],"class_list":["post-428772","post","type-post","status-publish","format-standard","has-post-thumbnail","category-technology","tag-ai","tag-artificial-intelligence","tag-chatgpt","tag-christoph-riedl","tag-il","tag-israel","tag-technology"],"_links":{"self":[{"href":"https:\/\/www.newsbeep.com\/il\/wp-json\/wp\/v2\/posts\/428772","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.newsbeep.com\/il\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.newsbeep.com\/il\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/il\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/il\/wp-json\/wp\/v2\/comments?post=428772"}],"version-history":[{"count":0,"href":"https:\/\/www.newsbeep.com\/il\/wp-json\/wp\/v2\/posts\/428772\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/il\/wp-json\/wp\/v2\/media\/428773"}],"wp:attachment":[{"href":"https:\/\/www.newsbeep.com\/il\/wp-json\/wp\/v2\/media?parent=428772"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.newsbeep.com\/il\/wp-json\/wp\/v2\/categories?post=428772"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.newsbeep.com\/il\/wp-json\/wp\/v2\/tags?post=428772"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}![\"Christoph](\"https:\/\/www.newsbeep.com\/il\/wp-content\/uploads\/2026\/05\/Christoph-Riedl.jpg\")
ChatGPT\u2019s goblin problem comes down to the reinforcement training used to fine tune AI models, said Christoph Riedl, a professor of computer science, information systems and network science at Northeastern University. Photo by Adam Glanzman\/Northeastern University<\/p>\n![\"A](\"https:\/\/www.newsbeep.com\/il\/wp-content\/uploads\/2026\/05\/Goblins-increased-in-GPT-5.4-especially-for-the-Nerdy-personality.png\")
The sharpest increase in goblin references occurred in responses from ChatGPT\u2019s \u201cnerdy\u201d personality profile. But there were also small but noticeable upticks in goblin appearances across the other custom personalities too. Credit: OpenAI<\/p>\n