Elon Musk’s artificial-intelligence company xAI has quietly made hundreds of thousands of conversations with its Grok chatbot publicly searchable, according to a Forbes investigation published on 20 August. Each time a user pressed Grok’s “share” button, the product generated a unique web address that was automatically indexed by Google, Bing and DuckDuckGo, leaving at least 370,000 chats visible to anyone on the Internet without warning to users. The indexed material ranges from mundane marketing requests to highly sensitive content. Forbes reviewed exchanges that revealed medical queries, passwords, spreadsheets, images and, in several cases, prohibited instructions on producing fentanyl, building explosive devices and carrying out an assassination plot against Musk himself. Some conversations also showed the bot ignoring xAI’s own safety rules. The disclosure follows a similar, short-lived incident at OpenAI in which ChatGPT transcripts appeared in search results before the rival firm disabled the feature. xAI, by contrast, has not responded to multiple requests for comment or indicated any policy change. Security researchers warn that the indexing loophole could be exploited for social-engineering attacks and search-engine manipulation while exposing users to privacy and legal risks.