OpenAI disabled access to ChatGPT for several hours after some users were able to see other users’ personal information and chats.

ChatGPTPhoto: Jonathan Raa / Zuma Press / Profimedia Images

Earlier this week, some ChatGPT users noticed that information and discussions that did not belong to them appeared in their accounts. To investigate and resolve the issue, OpenAI blocked access to the chatbot for most of Monday, News.ro reported.

Now the company is going to explain what happened and what data was actually released.

It was a cache issue, OpenAI explains. The open source tool used, Redis, returned different query data when one of the queries was canceled before completion.

In addition to the Redis bug, a big problem was the misconfiguration of OpenAI’s own servers, which caused a wave of canceled Redis requests.

Thus, the data of some users got into the accounts of other ChatGPT users. The company says only 1.2% of Premium subscribers were affected.

In this way, a whole series of personal information was compromised, from the name and email address to the expiration date of the card and its last four digits.

The same bug allowed parts of ChatGPT conversations to get into other users’ accounts.

  • ChatGPT had a “significant problem” / what OpenAI says was fixable?