Hundreds of Thousands of Grok Chatbot Conversations Leaked Online

The Grok chatbot, developed by Elon Musk’s artificial intelligence company xAI, has gained popularity due to its ability to perform real-time online searches and analyze social media. However, a recent incident has raised significant concerns about user privacy, as hundreds of thousands of user conversations with the Grok chatbot have been exposed online. These private conversations have become accessible through Google Search, sparking new worries about online privacy for Grok users.

The issue stems from Grok’s “share” button, which generates a unique URL for users to share specific conversations via email, text, or social media. However, these URLs were not restricted to private recipients and were instead indexed by search engines like Google, Bing, and DuckDuckGo. As of Thursday, approximately 370,000 Grok chatbot conversations were found to be publicly searchable. Technology experts have described this incident as a “privacy disaster.” The exposed conversations include advice on creating secure passwords, meal plans for weight loss, and responses to complex medical queries.

Experts note that while user names or account details may not be directly visible in shared chats, sensitive personal information could still be revealed, increasing privacy risks. Luc Rocher, Associate Professor at the Oxford Internet Institute, stated, “Leaked conversations include users’ names, locations, mental health details, business information, or personal relationships. Once online, this information remains there forever.” Carissa Veliz, Associate Professor at Oxford University’s Institute for Ethics in AI, added, “It’s concerning that users were not informed their chats would appear in search results. The biggest issue is that technology doesn’t tell us what’s being done with our data.” A similar data leak incident occurred with ChatGPT, where shared conversations were indexed by Google. OpenAI, the creator of ChatGPT, has already discontinued this experimental feature, citing risks of unintentional data exposure. According to OpenAI, the option to make ChatGPT conversations discoverable was a short-lived experiment and has since been removed.