Claude chats and artifacts exposed via Google search

Over the weekend, Reddit users noticed that search operators like "site:claude.ai/share" show hundreds of shared chats from Claude AI, including Artifacts—interactive mini-applications. Among the found materials were medical reports with patient data, clinical trial results with names, documents containing names and phone numbers of elementary school children, internal company documents, and employee reviews. The issue is related to the "share chat" feature, which creates a link for viewing the conversation; the interface warns that anyone with the link can view it. Anthropic told TechCrunch that links appear in search results only if they are published on a forum or social network, and when sent privately, they are not indexed. Anthropic representative Amy Rotherham explained that the company does not share chat directories with search engines, and the links are not guessable. By Monday, the issue was resolved, and test searches returned no results. In one case, according to Fortune, a chat labeled "shared by Anthropic" was found in which Claude created erotic content, violating usage policy. Google stated that it does not control which pages become public and provides websites with tools to manage indexing. Previously, last year, Forbes reported a similar issue with hundreds of Claude chats, and 404 Media reported the possibility of extracting approximately 100,000 public ChatGPT chats.