Publicly shared conversations with Anthropic’s Claude AI temporarily appeared in Google search results, allowing anyone to browse hundreds of chats ranging from creative writing prompts to more personal requests.
The issue came to light after a Reddit user discovered that specific Google searches would reveal a large number of shared Claude conversations.
While the chats weren’t private conversations that had been leaked, many users were unaware that search engines could also index conversations shared using Claude’s built-in sharing feature.
Public Claude chats included everything from coding help to becoming a mythical fox #
The discovery prompted users to search through the indexed conversations, uncovering an eclectic mix of requests.
Among the more unusual examples was one user asking Claude to help them become a nine-tailed fox from Asian mythology, with the AI outlining how someone could embody the creature through meditation, visualization, and roleplay.
Other publicly accessible chats included a lawyer asking if they needed to self-report a potential breach of conduct, a user creating a crypto wallet and exposing passwords, and conspiracy theories suggesting birds were professional rappers responsible for Jay-Z’s music.
Unsurprisingly, a “staggering” amount of erotica was discovered by users in the thread, alongside, somewhat ironically, drafts for cybersecurity projects.
By default, Claude conversations are private, and users must deliberately generate a share link before a conversation becomes publicly accessible. The controversy instead stems from shared pages being indexed by Google, making them discoverable through search results rather than only by people who had been given the link directly.
Google has since removed indexed Claude conversations from its search results, and the original search term for shared Claude chats now returns no results.