A weekend privacy scare saw shared Claude conversations and Artifacts, the interactive mini-apps built inside the chatbot, surfacing directly in Google search results.
Users discovered that a simple search operator such as “site:claude.ai/share” could pull up thousands of publicly shared chats that many likely assumed were private.
The exposed material was wide-ranging and at times highly sensitive. Reports pointed to medical records and clinical trial data carrying patient names, company files marked “internal use only,” children’s names and phone numbers, employee reviews with personal details, and code and work notes stored in Artifacts.
The issue was first flagged by a Reddit user on Saturday and then reported by 404 Media on Monday morning. By Monday afternoon, TechCrunch’s own testing indicated the exposure had already been fixed. A similar episode was documented by Forbes back in 2025.
Anthropic pushed back on the framing of the problem. Spokesperson Amie Rotherham said, “We give people control over sharing their Claude conversations publicly… These shareable links are not guessable or discoverable unless people choose to share them themselves.”
The episode is a reminder that anything you explicitly choose to share from an AI chatbot can potentially be indexed by search engines, so sensitive details are best kept out of shared links.
Source: TechCrunch.
Comments 0