Some ChatGPT users have noticed a strange behaviour when they are entering a search query into the Large Language Model (LLM)-powered chatbot. In a few instances, the chatbot has been spotted referring to the users by their name despite never being told so -- making the entire exchange rather "creepy" as some have pointed out on social media.
Users shared screenshots of their conversation with the chatbot, where it greeted them by their name unprovoked. "Insanely creepy. I hate it. Been trying to figure out how to turn it off," wrote Nick Donbos, a developer.
Other users echoed a similar experience, stating that it wasn't previously the case when using ChatGPT.
"I also find this a bit annoying! It also says odd things, like "Daniel is working through XYZ" when actually it's the one working through XYZ," said another user.
A third commented: "I hate it. Feels like a weird invasion of privacy."
Also Read | OpenAI's Reasoning Models Are "Hallucinating" And Creators Have No Idea Why: Report
ChatGPT remembers everything
Though it is unclear what prompted the change, some users pointed out that the recent ChatGPT upgraded "memory" feature could be responsible for it.
Last week, OpenAI introduced a new update to ChatGPT's Memory feature, under which the chatbot is able to remember everything a user has told it over the years. The company stated that the bot will leverage users' "preferences and interests to make it even more helpful for writing, getting advice, learning, and beyond".
In addition to the saved memories, ChatGPT can also reference users' past chats to deliver responses that feel noticeably more relevant and useful.
At the time, users expressed concerns over the potential of a chatbot retaining all the data and its possible misuse by the company. However, OpenAI CEO Sam Altman released a statement, clarifying that users can "opt out of this, or memory altogether" if they want.
