Anthropic’s Claude chatbot “shared” conversation links reportedly surfaced in Google search results, exposing users’ seemingly private chat histories—including content that included sensitive material—according to reporting on the issue’s appearance and subsequent fix. The exposure appears tied to Claude’s “share” feature, which generates a snapshot of a conversation at a unique web address intended for a recipient. A technique circulated on social media allowed shared chat directories and content to be discoverable through search queries, based on user reports. Anthropic said it does not provide chat directories or sitemaps to search engines, and that shareable links are only accessible when users choose to share them; the company also stated it had addressed the discovery method and that previously exposed links were no longer reachable through the technique. For universities and higher-education research centers adopting AI tools, the incident is an immediate reminder to recheck model-sharing workflows, data classification practices, and third-party prompt-sharing controls—especially when students or staff use AI outside sanctioned systems.