Hundreds of thousands of Grok chats exposed to Google results

Hundreds of thousands of user conversations with Elon Musk's AI (AI) chatbot Grok have been published in search engine results.

A unique link is created when Grok users press a button to share a conversation transcript, but not only do they share the chat with the intended recipient, but the button also appears to have made chat searchable online.

A Google search on Thursday revealed that around 300,000 GROK conversations were indexed.

This led to one expert to describe the AI ​​chatbot as an “ongoing privacy disaster.”

The BBC is approaching X for comments.

The arrival of Grok chat in Search Engine results was first reported by Tech Industry Publication Forbes, which counted over 370,000 user conversations on Google.

Among the chat transcripts seen on the BBC were examples of masked chatbots being asked to create a secure password, provide a weight loss meal plan, and answer detailed questions about their medical condition.

Some indexed transcripts have shown user attempts to test the limitations on what Grok says or does.

In one example seen in the BBC, the chatbot provided detailed instructions on how to make Class A drugs in the lab.

It's not the first time people's conversations with AI chatbots have been shown more broadly than they originally achieved when using the “sharing” feature.

Openai recently returned an “experiment” in which users saw ChatGPT conversations appear in search engine results when shared.

A spokesperson told BBC News at the time, “we were testing ways to make it easy to share useful conversations while still controlling users.”

They said that user chats are private by default and users must explicitly opt-in to share them.

Earlier this year, Meta faced criticism after sharing a conversation with Chatbot Meta AI in the app's “discovery” feed.

User account details may be anonymized or obscure in the shared chatbot transcript, but the prompt may still contain personal, sensitive information about someone.

Experts say it emphasizes that this increases privacy concerns among users.

“AI chatbots are an ongoing privacy disaster,” Professor Luc Rocher, an associate professor at the Oxford Internet Institute, told the BBC.

They said “leaked conversations” from chatbots fluctuated user information from full names and locations to sensitive details about mental health, business operations or relationships.

“When it's leaked online, these conversations stay there forever,” they added.

Meanwhile, Carissa Belis, an associate professor of philosophy at the Institute of Ethics at AI's Oxford University, said users who are not told that a shared chat would appear in search results “is a problem.”

“Our technology doesn't even tell us what it's doing with our data. That's a problem,” she said.