Anthropic said it addressed an issue in which a trove of “shared” Claude conversations appeared in Google search results, including instances containing sensitive information. The exposure was tied to Claude’s “share” feature, which creates a public snapshot intended for limited sharing. Users reportedly found that a specific search phrase could surface lists of shared chat directories, with some conversations including cryptocurrency wallet keys and personal identifiers such as names and addresses. Claude’s “Artifacts” also showed up in search results. Anthropic said it does not provide chat directories or sitemaps to search engines, arguing the content becomes public only when users choose to share it. Still, links reportedly remained reachable for some time after the issue surfaced, intensifying privacy and data-leak risk concerns for AI-enabled coursework and research tools.