Several Claude users’ private conversations using the shared chat and Artifacts feature were exposed after Redditors discovered it is searchable through Google.
The reveal raises fresh concerns about AI privacy and data security.
What happened?
Claude users can share URLs to their projects, allowing anyone with the link to view the conversation.
This feature’s issue came to light over the weekend when Reddit users discovered that typing “site:claude.ai/share” on Google searches displayed shared Claude conversations and Artifacts.
Some of the exposed content reportedly included medical records, internal company documents, and personal information such as children’s names and phone numbers.
Claude maker Anthropic stressed that conversations only appear in search results if users post those links on public forums and social media.
However, by Monday afternoon the searchable results had disappeared from Google, suggesting the issue had been addressed.
Reports claimed that sensitive information, private work notes, and software code had already been visible. One shared conversation also reportedly contained AI-generated explicit content, despite Anthropic’s policy banning sexually explicit material.
Not the first time
The incident echoes similar privacy concerns from last year, when hundreds of Claude chats and around 100,000 publicly shared ChatGPT conversations were reportedly indexed or scraped, highlighting ongoing risks around sharing AI-generated content online.