ChatGPT bug exposed data from subscribers and AI conversations

As announced by OpenAI this Friday (24), the temporary interruption to the operation of ChatGPT on Monday (20) was caused by a bug in the platform’s open source library (24). The problem also led to Disclosure of data from a small subscriber base of the paid version of the artificial intelligence (AI) tool.

According to the chatbot developer, the flaw allowed viewing the history of conversations with the AI ​​by other people. It didn’t take long for reports of the bug to surface on social media, resulting in the virtual robot being momentarily disabled.

Initially, the company believed that only third-party conversations were being accessed inappropriately. However, a more detailed study showed that Data from 1.2% of ChatGPT Plus subscribers was also exposedincluding full name, email address and payment information.

As for the leaked payment data, OpenAI says only the last four digits of the card registered by the user and the expiry date of the document could be displayed.. “Not all of the credit card numbers were ever disclosed,” confirmed the person in charge of ChatGPT.

The number of affected may be small

The startup, now one of the most valuable in the United States, said in a statement that the number of users whose data was actually viewed by others could be “extremely low.” In order to access the information, the system required some special conditions.

One was to open a confirmation email sent by ChatGPT to the wrong users during a nine-hour gap during Monday. The other was to go to the platform settings in the same time period, open the “My Account” menu and then go to the “Manage my subscription” option.

OpenAI also commented on the same error that led to: Data leak in ChatGPT, already fixed, it may have happened before, but no confirmed cases so far. Finally, the company announced that it has contacted affected users and informed them that their information may have been exposed.