Elon Musk’s artificial-intelligence venture xAI has inadvertently made hundreds of thousands of user conversations with its Grok chatbot publicly searchable on the open web, according to a Forbes investigation. Each time a user clicks Grok’s “share” button, the service generates a unique webpage that is automatically indexed by Google, Bing and DuckDuckGo. A Google query on 20 Aug. surfaced more than 370,000 Grok dialogues, many posted without the knowledge or consent of their authors. The exposed transcripts range from mundane marketing prompts to highly sensitive material, including passwords, personal health questions, uploaded files and instructions for manufacturing fentanyl, building bombs and assassinating Musk himself. Several chats appear to contravene Grok’s own usage policies, which ban content that promotes self-harm or facilitates violent or illicit activity. xAI has not responded to repeated requests for comment on how long the feature has been active or whether it will implement technical fixes such as no-index tags or explicit warnings. As recently as last month, the company asserted that Grok lacked any sharing function, and Musk publicly contrasted the product with OpenAI’s ChatGPT, which withdrew a similar tool after a backlash earlier this year. The incident intensifies scrutiny of privacy practices across the generative-AI sector and has renewed calls for comprehensive U.S. data-protection legislation. Harvard academics writing in Bloomberg Law on 21 Aug. argued that a federal privacy statute, including enforceable data-deletion rights, is needed to prevent exactly the kind of large-scale exposure now seen with Grok.