If OpenAI stated that one million netizens tested ChatGPT in the first week of its launch, we can imagine that to date there are closer to tens of millions using the intelligence-based conversation tool frequently. And inevitably, what had to happen happened: the service was interrupted on Monday after a bug was discovered, leaving certain users the possibility of partially accessing other people’s discussions.
ChatGPT saw a new lease of life last week with the arrival of GPT4, which aims to improve the machine’s capabilities and broaden the scope of the tool’s capabilities (e.g. using photos instead of text). The platform had experienced slowdowns as it grew in popularity, but no major bugs have ever been reported until now.
According to a spokesman for OpenAI, the company that developed ChatGPT, some internet users’ conversations were visible in other internet users’ chat history. As soon as the company became aware of the problem, it suspended the service while clarifying that the content of the conversations was not available to other accounts – only their titles. The error would come from the open source software used.
But at this point, OpenAI has not identified the reasons for this mixing of conversations between user accounts.
For the most part, the issue wasn’t that bothersome as it was still possible to use ChatGPT – and the title of chats isn’t usually a privacy issue (especially since external chats don’t mention the identity of their author).
As recently as yesterday, OpenAI paused chat history for everyone to avoid problems with information disclosure between accounts. All was back to normal in the US on Monday night when the OpenAI spokesperson said, “ChatGPT is back online, as is chat history.”
More information should follow in the next few hours so that OpenAI can at least confirm that it has completely fixed the bug at the source of this malfunction.