A ChatGPT bug exposed some users' conversation titles, it emerged overnight. Social media platforms, including Reddit and Twitter, witnessed users sharing images of chat histories that did not belong to them.
OpenAI CEO Sam Altman admitted the error was "significant", but assured that it had been resolved.
Since its launch in November last year, an estimated 100mln people have signed up to use ChatGPT for various tasks, including message drafting, songwriting and coding. Users can revisit their conversation history stored within the chatbot, but recently, some reported seeing unfamiliar conversations in their history.
One Reddit user shared a screenshot featuring chat titles like "Chinese Socialism Development" and Mandarin conversations.
In response to the issue, OpenAI told Bloomberg that they briefly disabled ChatGPT on Monday to address the glitch, adding that users could not access actual chats.
While Altman announced a forthcoming "technical postmortem", the incident has raised concerns about potential data privacy breaches.
It occurred just a day after Google introduced its chatbot Bard to beta testers and journalists.
As major players like Google and Microsoft, an OpenAI investor, compete in the AI tools market, the rapid development and release of new products have led to concerns about possible unintended consequences or harm resulting from such missteps.