A trove of seemingly private conversations with Anthropic’s Claude chatbot appeared in Google Search results recently, sparking alarm after users discovered sensitive information including cryptocurrency wallet keys and personal details exposed inadvertently. According to a report by Fortune, the issue was traced to Claude’s ‘share’ feature which enables users to generate a link to send a conversation to others. Anthropic said the company itself did not make chats searchable, instead blaming the way users shared links. “We give people control over sharing their Claude conversations publicly, and in keeping with our privacy principles, we do not share chat directories or sitemaps with search engines like Google,” a spokesperson told Fortune. The company emphasized that links are only accessible if users choose to share them, but once shared, they can be archived by third-party services. Anthropic has since fixed the issue, removing the links from Google search, though previously shared URLs remain accessible to anyone who already has them. The incident is not unique: last year, a similar glitch exposed nearly 100,000 ChatGPT conversations, while Elon Musk’s Grok chatbot also suffered from the same vulnerability.