How a Redis Client Bug Exposed ChatGPT Users’ Data – Inside the Leak

Incident Overview

On Monday, ChatGPT experienced a data‑leak incident where many users saw other users’ conversation histories, and some Plus users even observed other users’ email addresses on their subscription pages.

OpenAI temporarily disabled ChatGPT to investigate and CEO Sam Altman posted a tweet acknowledging a major issue caused by an open‑source library error.

Technical Details

The problem originated from a bug in the open‑source Redis client library redis‑py. OpenAI uses Redis to cache user information, employing Redis Cluster and an async connection pool via redis‑py for their Python servers.

When a request is cancelled after being queued but before a response is dequeued, the connection can become corrupted, causing subsequent unrelated requests to receive leftover data. This often results in unrecoverable server errors, but occasionally the corrupted data matches the expected type, leading to other users’ data being returned as valid.

On March 20 (Pacific Time), a change introduced a surge in cancelled Redis requests, increasing the chance of returning erroneous data. The bug was specific to the async Redis‑py client used with Redis Cluster and has now been fixed.

Impact and Response

OpenAI’s investigation revealed that a small number of users could see other active users’ names, email addresses, billing addresses, the last four digits of credit card numbers, and expiration dates. Full credit card numbers were not exposed.

This affected approximately 1.2% of ChatGPT Plus users, and OpenAI is contacting all impacted users.