Google Indexing Sensitive ChatGPT Chats, Exposing User Data

Google is indexing private conversations held on ChatGPT that users have shared, making these exchanges publicly discoverable through search engine results. When a user utilizes the “Share” button on a ChatGPT conversation, a unique link is generated. A basic Google site search, leveraging parts of these links, can reveal conversations originally intended for a small, private audience of friends, family, or colleagues.

These publicly accessible conversations have been found to contain highly sensitive personal information. Examples include individuals discussing struggles with addiction, experiences of physical abuse, serious mental health issues, and even expressing concerns about AI models potentially monitoring them. While ChatGPT itself does not display user identities, individuals may inadvertently reveal personally identifiable details within the chat content.

It is unclear whether users are fully aware that clicking the “Share” button, often intended for sharing with a select few via messaging platforms like WhatsApp or for personal record-keeping, renders their conversations publicly accessible and searchable.

Approximately 4,500 such conversations have already appeared in Google search results. While many of these do not contain deeply personal or identifying information, a significant number do. This figure likely represents only a fraction of the total conversations Google may have indexed. Due to the highly personal nature of some of the disclosed information, which can include names, locations, and specific personal circumstances, direct links to these conversations are not being provided.