Incidents · ChatGPT Bug Exposes Users' Chat History and Payment Information
AI INCIDENT

ChatGPT Bug Exposes Users' Chat History and Payment Information high

Date
March 20, 2023
Company
OpenAI
Product
ChatGPT
Category
Privacy
Severity
HIGH

What Happened

On March 20, 2023, ChatGPT users began reporting that they could see other users’ conversation titles in their sidebar. OpenAI took the service offline to investigate and discovered a bug in the open-source Redis client library (redis-py). Under specific conditions involving cancelled requests during a narrow time window, connections could return data belonging to another user. The bug exposed conversation titles from other users’ chat histories.

More critically, OpenAI discovered the same bug could have exposed payment-related information for approximately 1.2% of ChatGPT Plus subscribers who were active during a nine-hour window. Exposed data included first and last names, email addresses, payment addresses, credit card type, and the last four digits of card numbers. Full credit card numbers were never exposed.

Why It Matters

This was OpenAI’s first major data breach and it occurred at a time when ChatGPT was experiencing explosive growth, having reached 100 million users faster than any consumer application in history. The incident demonstrated that even companies at the forefront of AI development face mundane but serious infrastructure vulnerabilities. It raised questions about the security posture of rapidly scaling AI services and provided ammunition to regulators arguing for stricter AI oversight. Italy’s data protection authority cited this incident as partial justification for its subsequent ban of ChatGPT.

Lessons Learned

Rapid scaling introduces infrastructure risks that are separate from AI-specific risks. Open-source dependency vulnerabilities affect AI companies just as they affect any software company. Data isolation architecture must account for edge cases in concurrent connection handling. Companies handling sensitive user data need robust anomaly detection that can identify cross-user data leakage in real time. Transparency in incident response builds more trust than silence.

Current Status

OpenAI published a detailed post-mortem explaining the technical cause. The company patched the Redis library, added redundant checks against serving data across user boundaries, and conducted a broader security audit. The incident prompted OpenAI to invest more heavily in security infrastructure. It remains one of the most cited examples when regulators discuss AI platform security requirements.