Forbes reports that Google Search provides easy access to hundreds of thousands of conversations users have had with Elon Musk’s xAI chatbot Grok.
When a Grok user selects the “share” option on a conversation with the chatbot, a distinctive URL is generated for the user to share the conversation via email, text, or social media.
Forbes reports that search engines such as Google, Bing, and DuckDuckGo index those URLs, enabling anyone to search for those conversations on the internet.
The chats leaked by Grok, which were recently affected by a similar issue, provide a glimpse into the less-than-respectable desires of users.
These chats include questions about how to hack crypto wallets, dirty chats with an explicit AI persona, and requests for instructions on cooking meth. Users of Meta and OpenAI’s chatbots were also affected by this issue.
Even though xAI’s policies prohibit the use of its bot to “promote critically harming human life” or to develop “bioweapons, chemical weapons, or weapons of mass destruction,” users have continued to request Grok’s assistance with these matters.
Grok provided users with instructions on the production of fentanyl, a comprehensive plan for the assassination of Elon Musk, a list of various suicide methods, and suggestions for bomb construction, as evidenced by conversations that were made accessible by Google.
XAI did not promptly address a request for comment. Additionally, we inquired about the date on which xAI initiated the indexing of Grok conversations.
OpenAI described the recent occurrence of ChatGPT users’ conversations being indexed on Google as a “short-lived experiment.” Users raised the alarm late last month.
Grok clarified that it “prioritizes privacy” and does not have a “sharing feature” in a post that Musk quoted and tweeted with the phrase “Grok ftw.”