370,000 Grok Chats Exposed: Musk’s AI Hit by Massive Privacy Breach

Hundreds of thousands of private GROK chats can be found publicly on the Internet. Elon Musk’s AI company Xai has unintentionally released personal conversations, documents and sometimes explosive inquiries.

Privacy disaster discovered at Grok

What users held for private conversations with Elon Musk’s Ki-Chatbot Grok can now be found for everyone on Google and other search engines. More than 370,000 conversations with the chat bot were indexed on the Internet – apparently without the knowledge or approval of the users concerned. The problem was created by an apparently harmless function: the partial button of the chatbot.

A comprehensive examination of the US economic magazine Forbes the extent of the data breakdown brought to light. When GROK users click on the “share” button, the chat bot creates a clear URL that makes the conversation accessible to others. What many did not know: These links are automatically made available and indexed for search engines such as Google, Bing and Co., which makes the conversations publicly searing.

Sensitive data and illegal content

The content of the now public chats range from harmless inquiries to highly sensitive and problematic content. While some users only asked the bat to write tweets or to summarize texts, other conversations contained personal data such as names, passwords and medical information. Even uploaded documents, tables and pictures were accessible via the shared GROK pages. Particularly questionable: Under the published discussions, there are also inquiries about instructions for hacking crypto wallets, explicit chats and even asking for recipes for the production of methamphetamine.

This is in direct contradiction to Xai’s own guidelines, which prohibits the use of the bot to “promote critical damage to human life” or to develop “organic monkeys, chemical weapons or weapons of mass destruction”. In some cases, despite these guidelines, Grok has provided detailed instructions for the production of illegal drugs such as fentanyl and methamphetamine, gave instructions for creating malware or building bombs and even listed methods for suicide. Particularly bizarre: In a chat, Grok even offered a detailed plan for the murder of Elon Musk himself.

Musks embarrassing U -turn

The unveiling is particularly piquant because Musk had recently criticized Openai for a similar problem. When Chatgpt users noticed a comparable data breakdown in July, the official GROK account on X (formerly Twitter) claimed that GROK “had no part” and “prioritized privacy”. Even experts were surprised by the problem.

Nathan Lambert, a computer scientist at all institutes for AI, used Grok to create summaries of his blog posts. He was shocked when he learned that his GROK request and the answer from the AI were indexed on Google. “I was surprised that GROK chats, which I shared with my team, were automatically indexed on Google, without any warning notices, especially after the recent turmoil with chatgpt,” said the researcher based in Seattle.

Responsibility and consequences

Google rejects responsibility and explains that website operators have full control over whether their content is indexed. “The editors of these pages have full control over whether they are indexed,” said a Google spokesman. This statement clearly makes responsibility at XAI, which has not yet responded to repeated inquiries for a statement.