Hundreds of thousands of conversations with Elon Musk’s Grok chatbot appeared in search engine results after users shared transcripts via unique links, making the chats accessible more widely than intended. According to BBC News, a Google search on Thursday showed nearly 300,000 indexed Grok conversations, after a share button generated URLs that were subsequently discoverable online. The BBC said it had approached X for comment.
Shared links made chats searchable
BBC News reports that Grok’s share feature creates unique links for conversation transcripts, which appear to have been indexed by Google, surfacing user chats publicly. Tech publication Forbes first reported the appearance of Grok chats in search results and counted more than 370,000 conversations indexed on Google, the BBC notes. Among transcripts seen by the BBC were prompts asking Grok to generate secure passwords, offer weight-loss meal plans, and answer detailed questions about medical conditions. Some users also appeared to test the system’s boundaries.
In an example cited by the BBC, the chatbot provided detailed instructions on how to make a Class A drug in a lab. The BBC highlights that this is not the first instance of shared AI chatbot conversations appearing more broadly via “share” functions. OpenAI previously rolled back an “experiment” that allowed shared ChatGPT conversations to show up in search engine results, saying at the time that chats were private by default and users had to explicitly opt in to sharing. Earlier this year, Meta faced criticism after shared conversations with its Meta AI appeared in a public “discover” feed on its app, the BBC reports.
Mounting privacy concerns
Experts warn shared prompts may reveal sensitive data
While shared transcripts can obscure account details, BBC News reports that prompts themselves may include personal or sensitive information. Prof Luc Rocher, associate professor at the Oxford Internet Institute, told the BBC that “AI chatbots are a privacy disaster in progress,” adding that “leaked conversations” have exposed details such as full names, locations, mental health information, business operations, or relationships, and that “once leaked online, these conversations will stay there forever.”
Carissa Veliz, associate professor in philosophy at Oxford University’s Institute for Ethics in AI, told the BBC that users not being told that shared chats would appear in search results is “problematic,” noting, “Our technology doesn’t even tell us what it’s doing with our data, and that’s a problem.” The BBC says it has contacted X regarding the Grok chat indexing and is awaiting a response.
