Reddit users found that shared Claude AI chats could be found through ordinary Google searches,
exposing private material such as crypto wallet keys, names, addresses, and work notes. The problem
came from Claude’s Share feature, which turns a conversation into a public web page that search
engines can index. Anthropic appears to have stopped Google from listing the pages, but links shared
earlier stay live and the same content is still findable on Bing.
1. What happened
Reddit users found a specific Google search query that surfaced conversations people had shared from Claude, Anthropic’s AI assistant. The exposed chats included crypto wallet keys, real names, home addresses, internal work notes, and personal or policy-violating content. The same query also surfaced Claude Artifacts, the interactive documents and apps that Claude can generate inside a conversation.
2. Why it happened
The exposure traces to Claude’s Share feature. Sharing a chat does not send it privately, it publishes a snapshot of the conversation at a public web address. Once search engines discover that address, the page behaves like any other public web content and can appear in results. Claude chats stay private by default, and only conversations a user chose to share were affected.
3. Current status
Anthropic appears to have fixed the Google indexing, according to Fortune. Links shared earlier are still live for anyone who already has the URL, so revoking them is a manual step. Wired reports the shared content is still findable through Bing.
- Grok Shared xAI Grok conversations appeared in Google results in August 2025.
- Meta AI Meta AI conversations could become public by design, tied to the user’s social account.
- ChatGPT OpenAI removed a share experiment in 2025 after users unintentionally made thousands of chats searchable.
4. Recommendations
- In Claude, open Settings, then Privacy, then Shared chats to review every shared conversation and revoke any link that should not be public.
- Do not paste personally identifiable information, credentials, wallet keys, or sensitive business data into any AI chatbot, and treat the share button as public publishing, not private messaging.
- Set a policy on which AI tools staff may use, and prefer temporary or incognito chat modes that are not saved to history for sensitive work.
5. Sources
- https://www.malwarebytes.com/blog/privacy/2026/07/shared-claude-chats-were-searchable-on-go
ogle - https://fortune.com/2026/07/27/a-trove-of-users-seemingly-private-conversations-with-anthropics-claude-ai-chatbot-showed-up-in-google-search-results
- https://www.wired.com/story/private-claude-chats-exposed-in-google-and-bing-search-results/
Enjoyed reading this Threat Intelligence Advisory? Stay updated with our latest exclusive content by following us on Twitter and LinkedIn
No related posts found.
