On March 20, 2023, some users of OpenAI’s ChatGPT service reported seeing the titles of other users’ conversation histories in their own accounts. The incident prompted OpenAI to temporarily take the AI chatbot offline to investigate and address the data leak.

In a post-mortem, OpenAI confirmed that the bug briefly exposed sensitive user data. The company attributed the issue to a flaw in an open-source library, redis-py, which it uses for caching user information in its servers.

Details of the Data Exposure

The bug allowed some users to see the titles of another active user’s conversation history. After fixing the initial bug, OpenAI discovered that the same flaw had a more serious implication. It caused the unintentional visibility of payment-related information for 1.2% of ChatGPT Plus subscribers during a specific nine-hour window on March 20.

The exposed information for this subset of subscribers could include the user’s first and last name, email address, payment address, the last four digits of their credit card number, and their credit card’s expiration date. OpenAI clarified that full credit card numbers were not exposed at any time.

OpenAI’s Response and Fix

OpenAI CEO Sam Altman acknowledged the incident, calling it a “significant issue” and confirmed that the underlying bug had been fixed. The company stated that the number of users whose payment data was actually revealed to another person was “extremely low.” OpenAI followed up by notifying the users whose payment information may have been exposed due to the bug. The service was restored after the patch was implemented.