Claude users are being urged to review conversations they have shared after reports that some publicly shared chats and Artifacts briefly appeared in search engine results, raising fresh concerns about privacy on AI platforms.
The issue gained widespread attention after a Reddit discussion attracted thousands of upvotes, with users claiming that Claude share links could be discovered through search engines using queries such as site:claude.ai/share.
Some of the indexed pages reportedly contained resumes, software code, legal discussions, API keys and cryptocurrency wallet information.
The controversy centres on how many users interpreted Claude’s sharing feature. While Anthropic labels shared conversations as accessible to “anyone with the link,” many users assumed those links would remain effectively unlisted rather than potentially discoverable through search engines.
Anthropic’s own documentation states that creating a share link generates a public snapshot of a conversation that anyone with the link can view.
Developer Om Patel, who helped draw attention to the issue on X, claimed the problem extended beyond conversations to Claude Artifacts, including interactive apps, dashboards and documents. However, some of those specific claims have not been independently verified.
By Monday, users reported that Google was no longer returning the shared pages through the same search queries, suggesting the issue had been addressed or the pages had been removed from Google’s index.
In a statement to Fortune, Anthropic said it does not provide chat directories or sitemaps to search engines and stressed that users deliberately make conversations public when they create share links.
“When someone shares a conversation, they are making that content publicly accessible, and like other public web content, it may be archived by third-party services,” an Anthropic spokesperson said.
The incident has renewed discussion about how AI platforms handle shared content. While private Claude conversations were not breached or exposed, cybersecurity experts say anything published through a public URL should be treated as potentially discoverable, even if it is intended for a limited audience.
Users who have previously shared Claude conversations or Artifacts can review and revoke those links from their Claude account settings.
Security experts also recommend avoiding the inclusion of passwords, API keys, confidential business information or other sensitive personal data in any AI conversation that is shared publicly.
