OpenAI Patches ChatGPT Bug Exposing User Conversation Titles
OpenAI resolves a glitch that allowed a small number of ChatGPT users to view the titles of other users' chat histories. The company confirms that while the conversation headings were visible, the actual content of the messages remained private.
OpenAI fixes a significant bug in ChatGPT that accidentally exposes the titles of other users' conversation histories. Chief Executive Sam Altman publicly apologizes for the glitch, which stems from an error in an open-source library, and notes that users temporarily lose access to their own chat logs on March 20 as a result of the patch.
The company clarifies that while the conversation headings appear in the sidebar, the actual substance of other users' chats remains completely secure and invisible. This automated labeling system typically generates a tab name based on the first query of a new session, and those generated titles are the only pieces of data that slip through the privacy flaw.
This incident occurs during a period of meteoric growth for the viral chatbot, just days after OpenAI launches its advanced GPT-4 model. As millions of people integrate ChatGPT into daily tasks like coding, writing, and designing, Altman promises a thorough technical postmortem to prevent similar privacy issues from happening in the future.