A TechCrunch report flagged on Hacker News shows that some publicly shared Claude chats and Artifacts have appeared in Google search results.
The discussion drew 18 points and 7 comments. Users reported finding their own conversations indexed despite intending limited sharing.
What Happened
Anthropic's share feature creates public links for chats and Artifacts. These links became crawlable by search engines.
No confirmation exists on how many chats were indexed or for how long the exposure lasted.
How Claude Sharing Works
Users toggle a share setting inside Claude projects or individual chats. The resulting URL is accessible to anyone with the link.
Once shared, the content sits on Anthropic servers without additional robots.txt restrictions mentioned in the thread.
Community Reaction
HN commenters noted surprise that shared links lacked no-index headers. Several users discovered their test prompts and internal notes in search results.
One thread highlighted similar past issues with other AI platforms where public links were indexed within days.
Privacy Implications
Shared chats can contain API keys, code snippets, or proprietary prompts. Indexing turns temporary shares into permanent records.
The 7 comments focused on lack of clear warnings before generating a shareable link.
How to Check and Secure Your Chats
- Search Google for your Claude username or project names.
- Revoke public links inside the Claude interface immediately.
- Avoid sharing chats that contain sensitive data or internal reasoning.
Anthropic has not published an official remediation timeline.
Alternatives for Private Sharing
Tools like private Notion pages or self-hosted Markdown viewers keep content off public indexes. Local LLM interfaces such as Ollama or LM Studio offer zero external sharing by default.
| Platform | Default Indexing Risk | Share Control |
|---|---|---|
| Claude | High | Public link only |
| ChatGPT | Medium | Link + password option |
| Local Ollama | None | No external links |
Who Should Use This
Developers handling proprietary prompts or client data should disable sharing entirely. Researchers needing public examples can use dedicated throwaway accounts.
Teams requiring audit trails should export chats to internal wikis instead of relying on Anthropic links.
Bottom line: Public Claude shares carry an indexing risk that users must manually mitigate until Anthropic adds stronger defaults.
Early reports suggest the issue stems from missing crawler directives rather than intentional design.
Top comments (0)