Elon Musk’s Grok chatbot talks about Musk assassination, terrorist attacks, drug making in leaked private chats by xAI

Elon Musk’s artificial intelligence company, xAI has published the chat transcripts of over 370,000 conversations between users and its Grok chatbot, often without users’ explicit consent or awareness. These conversations include everything from mundane tasks to highly sensitive content, raising serious privacy, ethical, and security concerns. Search engines such as Google, Bing, and DuckDuckGo have indexed these transcripts, making them publicly searchable online, which has alarmed privacy advocates.

What is Grok chatbot?

Grok is a generative AI chatbot developed by Musk’s startup xAI, launched in late 2023 as a competitor in the rapidly evolving AI assistant market. The chatbot integrates with Musk’s social media platform X (formerly Twitter) and Tesla vehicles and has evolved through several versions, with Grok 3 and Grok 4 being the latest iterations featuring advanced reasoning capabilities and extensive training using powerful computing infrastructure. The bot earned its name from a science fiction term meaning deep understanding, reflecting xAI’s ambition to create a “maximum truth-seeking AI”.

How were the conversations published?

The issue stems from Grok’s “share” feature, which generates a unique URL whenever users choose to share a chat transcript. While this feature allows users to share conversations easily, the links were automatically published on Grok’s website and accessible to major search engines without users' knowledge or explicit informed consent. As a result, over 370,000 chat transcripts became publicly discoverable and searchable, including dialogues involving personal details, passwords, medical and psychological queries, business information, and highly sensitive instructions.


Forbes investigation uncovered troubling content indexed online, including instructions on manufacturing Class A drugs, assassination plots targeting Elon Musk, discussions surrounding fictitious terrorist incidents, and attempts to breach cryptocurrency wallets. These examples not only violate xAI’s own terms of service, which prohibit use cases that critically harm human life but also highlight how harmful or dangerous information can be disseminated via AI if not properly monitored and secured.

Response and implications

The publication of these conversations has led to widespread criticism of xAI’s data handling practices. Experts argue that the breach demonstrates the challenges inherent in balancing user convenience with privacy protections in AI systems. Similar issues have surfaced with other AI providers, including OpenAI, which briefly tested a comparable feature allowing users to share ChatGPT conversations, some of which also appeared in search engine results.

xAI has not publicly detailed steps taken to address this privacy lapse. The incident raises urgent questions about data security, consent, and AI governance at Musk’s company.

FAQ

Q. How did users' conversations become public?
When users clicked Grok’s “share” button, a unique URL for the transcript was created and automatically published on the Grok website without clear notice that the content would be indexed by search engines.

Q. Were users informed their chats would be publicly searchable?
No. Users were generally unaware that sharing a conversation would make it searchable and publicly accessible online, raising concerns about informed consent.

Q. What types of information were exposed?
Exposed information ranged from everyday queries to highly sensitive medical questions, personal details, passwords, instructions for harmful activities, and controversial or illegal content.