OpenAI Patches ChatGPT Bug That Exposed User Chat Titles
OpenAI fixes a glitch that allowed a small number of ChatGPT users to view the titles of other users' conversation histories. The company confirms that the actual content of the chats remained private and is conducting a technical postmortem.
OpenAI resolves a significant bug in ChatGPT that accidentally exposes the titles of other users' conversation histories to a small group of people. Chief Executive Sam Altman attributes the glitch to an open-source library and states that the company feels awful about the oversight. Although the conversation titles appear in the sidebar, the actual substance of the chats remains completely hidden from other users.
To implement the fix, OpenAI temporarily disables the chat history feature on March 20, leaving many users unable to access their past conversations. The company automatically generates these visible titles based on the subject of the first query in any given chat session. Altman promises a thorough technical postmortem to explain exactly how the privacy slip occurred and how the team prevents similar issues.
This incident happens during a period of meteoric growth for the viral chatbot, which millions of people currently use for coding, writing, and design tasks. It also follows the recent launch of the upgraded GPT-4 model and the integration of OpenAI's technology into Microsoft's Bing search engine. As the artificial intelligence arms race accelerates across the technology sector, maintaining user trust and data privacy remains a critical challenge for OpenAI.