Hundreds of Claude AI chats unintentionally exposed in Google search results

A large number of conversations shared through Anthropic’s Claude AI assistant were found indexed by Google after a Reddit user demonstrated that a simple site‑specific search (site:claude.ai/share) returned publicly accessible pages. The exposed pages contained a range of content, including legal strategies, source code, private business communications, authentication keys, CVs, and cryptocurrency wallet details. The indexing issue stemmed from a conflict between the site’s robots.txt file, which blocks crawlers, and a “noindex” directive that should have removed the pages from search results. The same problem also affected Microsoft’s Bing search engine.

Most of the pages were removed from Google’s results by Sunday, according to reports from security‑focused sites, though Anthropic had not yet issued an official comment. This incident follows a similar exposure in September 2025 that affected roughly 600 Claude conversations, indicating a recurring privacy vulnerability in the service’s sharing feature.