OpenAI, the company behind ChatGPT, recently revealed that a software vulnerability caused the service to fail for several hours on Monday, exposing the addresses of users’ conversations. The company deliberately disabled the service to fix the vulnerability which led to the random appearance of addresses for users’ chat history with ChatGPT in other users. However, OpenAI confirmed that the error only led to the disclosure of the addresses of the conversations, and not to the leakage of the conversations themselves.
The company became aware of the issue after some ChatGPT users posted screenshots on social networks showing titles of conversations they didn’t have themselves appearing in the app’s sidebar. OpenAI was quick to temporarily close the service to preserve users’ privacy and to work on resolving the issue. Although the company did not reveal any technical details about the cause of the problem, it assured users that they could return to their previous chat history through the sidebar shown on the home page of the site, as the service automatically generates a title for each conversation to facilitate returning to it later.
On Monday, many ChatGPT users experienced difficulties accessing the site or encountered slow processing requests. The glitch affected both free and paid service subscribers, including ChatGPT Plus subscribers who expressed their dissatisfaction with the lack of access to the service or the slow response, which lasted for several hours. Users from around the world shared screenshots and images showing the service not working, and they expressed their frustration on social media.
It’s worth noting that OpenAI recently revealed the new artificial intelligence model GPT-4, which is a significant development of the GPT-3.5 model. The company also released a version of ChatGPT based on the new model, which is more accurate and intelligent than the previous version. However, the new version is still limited to paid service subscribers from ChatGPT.