Home / Technology / AI Privacy Glitch: Shared Claude Chats Exposed Online
AI Privacy Glitch: Shared Claude Chats Exposed Online
27 Jul
Summary
- Shared Claude AI conversations were publicly accessible via Google Search.
- A feature allowing users to share chats inadvertently made them indexable.
- Anthropic addressed the issue, and Google has removed the exposed chats.
Shared conversations from Anthropic's AI chatbot, Claude, were recently found to be publicly accessible through Google Search. This discovery, made by Reddit users, revealed that chats shared by Claude users were being indexed, exposing potentially sensitive information. For instance, one shared chat exposed cryptocurrency wallet keys, while another involved a lawyer seeking advice on a breach of conduct.
The issue stemmed from Claude's chat-sharing feature. While conversations are private by default, the static pages created for sharing lacked proper 'noindex' tags, allowing search engines like Google to index them. Users were able to find these conversations by searching specific query patterns on Google.
Anthropic has since addressed the problem, and Google has reportedly removed the indexed shared chats. This incident echoes similar privacy concerns with AI chatbots, including past instances with both ChatGPT and Claude itself, where shared conversations were previously found in search results.
