Most of us expect a reasonable degree of privacy when conversing with AI models like ChatGPT or Grok . While we understand that our interactions can be used to train and improve these models, you wouldn’t necessarily expect these conversations to appear in search, right? Unfortunately for Grok users, it looks like some of your chats are appearing on Google .

Grok chats are showing up on Google

If you were to do a search for Grok chats on Google, you would find as many as 370,000 user conversations. These conversations cover a variety of topics, including writing tweets, generating images, and more. Some of these conversations are also intimate in nature, like medicine and psychology. In some cases, names and personal details that users shared with Grok also show up.

If this situation sounds familiar, it is because recently, users of OpenAI’s ChatGPT encountered a similar problem . However, OpenAI claimed back then that it was a public sharing “experiment.” However, in Grok’s case, there was no warning or indication that this could happen.

This occurs when users click the “Share” button in their chats and conversations with Grok. However, most of us would assume that when we click “Share”, we plan to either share it with ourselves or the target recipient, not with the entire world or be indexed on Google. It is unclear if this was a bug or if it was intended, but either way, users were not warned ahead of time. xAI, Grok’s parent company, has remained silent.

Violation of xAI’s policies

The content of some of these conversations is a little embarrassing. However, the bigger issue here is that some of these chats are a clear violation of xAI’s rules . In addition to generating images and tweets, some of the topics discussed include requesting for instructions on hacking crypto wallets, manufacturing drugs, suicide methods, weapon creation, and more. There was even one involving a detailed assassination plan targeting xAI’s CEO, Elon Musk.

This suggests that while they go against the company’s rules, there doesn’t appear to be any kind of enforcement .