Thousands of Grok chats now visible on Google search
The issue stems from Grok’s “share” button, which allows users to generate a unique URL to post conversations via email, text, or social media.
ADVERTISEMENT
Hundreds of thousands of conversations with Elon Musk's artificial intelligence chatbot Grok were discovered freely accessible through Google Search, according to a Forbes report.
The leak has once again highlighted privacy concerns surrounding AI chatbots, after similar incidents recently affected OpenAI's ChatGPT and Meta's AI systems.
The issue stems from Grok's "share" button, which allows users to generate a unique URL to post conversations via email, text, or social media. Those URLs, however, are being indexed by search engines such as Google, Bing, and DuckDuckGo, making them searchable and accessible to anyone online, according to the report.
Much like previous chatbot leaks, the indexed Grok chats reveal a troubling side of user behaviour. Some of the exposed conversations include queries about hacking crypto wallets, sexually explicit exchanges with AI personas, and even requests for instructions on cooking meth.
xAI's usage rules explicitly prohibit the chatbot from promoting harm, developing chemical or biological weapons, or encouraging violence. However, as the leaked chats suggest, many users attempted to push Grok into generating such content regardless of the rules in place.
Just weeks ago, OpenAI faced flak after some ChatGPT chats surfaced on Google. At the same time, OpenAI called it a "short-lived experiment." Musk shared a post on X claiming that Grok had "no such sharing feature" and that it "prioritizes privacy."
As of now, xAI has not responded to requests for comments on when Grok chats began being indexed, the report added.
