Claude Shared Chats and Artifacts Found Publicly Searchable on Google
Over the weekend, Reddit users discovered that numerous shared chats and interactive projects (Artifacts) from Anthropic’s AI chatbot Claude were publicly accessible via Google search. By using operators like 'site:claude.ai/share', they found a range of sensitive content including health records, private company documents, children's personal information, code notes, employee reviews, and even explicit material generated by the bot. Anthropic attributed the exposure to users who shared links publicly rather than any system flaw. The company stated that shareable links only appear in search results if posted on public platforms like forums or social media. However, this explanation has been met with skepticism given the sensitive nature of some exposed content and a reported violation of Anthropic’s policy against generating explicit material. The issue was first flagged by a Reddit user on Saturday and later confirmed by multiple outlets including 404 Media and Futurism. As of Monday afternoon, TechCrunch found no results using similar search methods, suggesting the problem may have been resolved.
Topics: AI platforms data privacy security flaw