Anthropic Claude Data Exposure Incident
Numerous conversations with the Anthropic Claude chatbot, alongside data from its Artifacts mini-applications, have been discovered publicly accessible via Google search results. This incident raises significant concerns regarding user data privacy.
Origin of the Issue
The problem appears to stem from Claude’s “share chat” feature. This functionality is designed to allow users to generate links, providing anyone with the assigned URL the ability to view a specific conversation or project.
Sensitive Information Discovered
Among the publicly exposed data, highly sensitive information was found, including:
- Medical records
- Confidential company documents
- Names and phone numbers of minors
These findings underscore the critical importance of carefully reviewing privacy settings when utilizing AI tools that offer sharing capabilities.
Wow, this is a huge wake-up call. I’ve been using Claude for brainstorming and even some early-stage code snippets, and I definitely relied on the ‘share chat’ feature for quick collaboration with my team. I always assumed it was a private link unless I actively posted it somewhere. The idea of medical records and company documents being out there is terrifying. I’ll be scrubbing my shared chats and sticking to local copies for sensitive info from now on. My tip: always treat shared links as potentially public, no matter what the tool claims.