Elon Musk's Grok AI chatbot has reportedly released the chats of lakhs of users on the internet. The report claims that the AI chatbot has exposed the conversations of more than 370,000 users, which are now publicly available on Google. According to a report by Forbes, sensitive information, including medical and psychological questions, business details, and at least one password, was indexed by Google Chrome and other search engines. The report explains that the data exposure may be linked to Grok's "share" feature, which creates a unique, publicly accessible URL for each shared conversation. These links were published on Grok's website and left open to search engines, apparently without users' knowledge.
Multiple such transcripts that were reportedly indexed on Google Search have violated Grok’s terms of service. In one of the chats, Grok explained how to manufacture a Class A drug, while another gave detailed guidance on assassinating Elon Musk. xAI ’s rules explicitly ban using Grok for activities that “critically harm human life.”
OpenAI’s experiment with a similar feature backfired
Apart from Musk’s xAI, other AI companies have also faced similar issues with chatbot conversations appearing publicly. Earlier this month, some OpenAI ChatGPT users noticed their chats showing up in Google search results, even though they had chosen to make them “discoverable.”
Following user concerns, OpenAI ended the practice. Chief information security officer Dane Stuckey described it on X as a “short-lived experiment” that risked people unintentionally sharing information they did not want to.
After OpenAI scrapped its share feature, Musk celebrated the move. At the time, Grok’s X account stated it didn’t have such a feature, and Musk responded with a tweet saying, “Grok ftw” [for the win].
The timeline of when Grok introduced the sharing option remains unclear, but X users have been cautioning since January that Grok conversations were appearing in Google search results, the report adds.
Multiple such transcripts that were reportedly indexed on Google Search have violated Grok’s terms of service. In one of the chats, Grok explained how to manufacture a Class A drug, while another gave detailed guidance on assassinating Elon Musk. xAI ’s rules explicitly ban using Grok for activities that “critically harm human life.”
OpenAI’s experiment with a similar feature backfired
Apart from Musk’s xAI, other AI companies have also faced similar issues with chatbot conversations appearing publicly. Earlier this month, some OpenAI ChatGPT users noticed their chats showing up in Google search results, even though they had chosen to make them “discoverable.”
Following user concerns, OpenAI ended the practice. Chief information security officer Dane Stuckey described it on X as a “short-lived experiment” that risked people unintentionally sharing information they did not want to.
After OpenAI scrapped its share feature, Musk celebrated the move. At the time, Grok’s X account stated it didn’t have such a feature, and Musk responded with a tweet saying, “Grok ftw” [for the win].
The timeline of when Grok introduced the sharing option remains unclear, but X users have been cautioning since January that Grok conversations were appearing in Google search results, the report adds.
You may also like
Tories demand crackdown on people playing loud music on public transport
Lesser-known 'masterpiece' on Amazon Prime is 'mix of war film and gangster thriller'
2.2 million sign petition for Indian-origin trucker Harjinder Singh: Support grows for leniency; 45-year sentence looms
Gabriel Jesus drops new Arsenal injury update amid Mikel Arteta return boost
Eberechi Eze explains Tottenham snub in first comments on £67.5m Arsenal transfer
