Key facts
- Hundreds of Claude conversations were discoverable via Google search.
- Sensitive data including crypto wallet credentials and Social Security numbers were exposed.
- The root cause was a missing noindex meta tag on Anthropic's link-sharing feature.
- Anthropic has since updated its configuration to prevent further indexing by search engines.
- Some shared links were still being surfaced by Bing.
- A GitHub repository has archived a significant number of the exposed conversations.
Hundreds of conversations shared via Anthropic's Claude chatbot were inadvertently published and indexed by Google, exposing sensitive user data. A Reddit user discovered the issue on July 25, finding that a simple Google search returned readily accessible chats containing crypto wallet credentials, apparent Social Security numbers, legal discussions, and personal content.
The root cause was identified as a missing noindex meta tag on Anthropic's link-sharing feature. While Anthropic's robots.txt file was in place, it could not prevent indexing once share links appeared on public websites. This allowed search engines like Google to index pages that were intended to be private.
Similar to a past incident with OpenAI's ChatGPT, where a "Make this chat discoverable" option led to public indexing, Claude's share links became fully indexed webpages. The issue also affected Claude's published artifacts, such as payroll spreadsheets and internal customer relationship management exports.
Anthropic has since updated its configuration by adding noindex tags, and Google began removing the exposed Claude results by July 26. However, Bing was still surfacing shared links at the time of publication. A GitHub repository has already archived hundreds of the exposed conversations, with over 11,000 messages saved in plain text.