(Credit: VINCENT FEURAY/Hans Lucas/AFP via Getty Images)
OpenAI's ChatGPT has started referring to users by their names without being asked, which some find off-putting.
According to screencaps posted by one software developer on X, first spotted by TechCrunch, ChatGPT used his real first name when asked about a math problem involving integers, which he called it “creepy and unnecessary.”
Another user took things one step further, calling it “insanely creepy,” adding, “I hate it.”
Some users also noted that ChatGPT is using their names, despite not recalling ever giving ChatGPT that information.
Granted, it appears the quirk only impacts certain GPT variations like OpenAI o3, which are trained to “think longer before responding” and show their chain of thought after you ask a question. The “creepy” naming only appears on the screen when ChatGPT is showing its chain of thought and not in the official answer.
OpenAI has not yet released a statement about the quirk. However, TechCrunch noted that these complaints surfaced shortly after OpenAI rolled out a new feature earlier this month that enabled the chatbot to draw from previous conversations it had with users when formulating its answers. After announcing the feature, OpenAI CEO Sam Altman said AI systems will "get to know you over your life, and become extremely useful and personalized."
Though being named by ChatGPT may understandably be a little weird, it's not the most unsettling surprise output we’ve seen from a mainstream large language models (LLM) recently. Google Gemini allegedly told one grad student to “Please die” and called them a “stain on the universe,” after being asked a simple question for gerontology coursework. Meanwhile, start-up Character.AI is now facing a lawsuit after being accused of playing a role in the suicide of a 14-year-old boy in Florida.