Claude chats exposed in Google searches – another endorsement of Apple’s privacy approach - 9to5Mac

What the report says

9to5Mac reported that some shared conversations from Anthropic’s Claude chatbot were discoverable through Google search, raising fresh concerns about the privacy risks of AI chat tools. The article, published July 28, 2026, said Reddit users found that searches targeting Claude’s shared-chat directory could surface conversations that users may have expected to be accessible only through a specific link.

According to 9to5Mac, citing BBC News, more than 200 Claude conversations appeared across at least 25 pages of search results. The exposed material reportedly included benign or odd prompts, but also sensitive content such as private conversation transcripts, a legal professional asking about a possible conduct-reporting obligation, and a user seeking help with a crypto wallet while revealing keys. Anthropic’s position, as quoted in the article, was that shared conversations are publicly accessible and may be archived by third-party services.

The article also cited Wired’s finding that sampled shared Claude pages lacked a “noindex” tag, a signal search engines consider when deciding whether to index pages. 9to5Mac said the search results appeared to stop working after the issue drew attention, but advised users not to rely on shared-chat links for confidentiality and suggested reviewing or deleting past shared chats through Claude’s privacy settings.

9to5Mac framed the incident as support for Apple’s privacy-focused approach to AI. The article contrasted Claude’s shared-chat exposure with Apple Intelligence and Siri’s stated model: handling requests on device where possible, using Apple’s Private Cloud Compute as a fallback, and applying similar protections when outside AI models are used. The broader takeaway is that AI sharing features can create privacy exposure if users misunderstand how widely content may be reachable.