Claude chats appeared in Google search after users used share links

Claude chats appeared in Google search results over the weekend after users had used Anthropic’s sharing tool, Fortune reported. The exposure matters because some of the conversations reportedly contained sensitive material, including cryptocurrency wallet keys, names and addresses.

The affected conversations were not ordinary private chats inside Claude accounts, according to Fortune. They were chats that users had turned into shareable web links, a feature meant to let someone send a conversation record to another person.

Members of a Reddit discussion channel focused on Claude noticed the issue when a particular Google search query returned many shared Claude conversations, Fortune reported. Claude Artifacts, the interactive tools and small apps users can build inside the chatbot, also appeared in the results.

Why did Claude chats appear in Google search?

AI chatbot share tools create a unique public URL for a conversation, which can be sent to other people. Fortune reported that in this case, shared Claude links were visible through Google search, even though users may not have expected the links to be discoverable that way.

The material found through search covered a wide range of topics, according to Fortune, including erotica, programming conversations, work notes and fake book reviews. One conversation marked as “shared by Anthropic” showed Claude producing explicit content, Fortune reported, which conflicts with Anthropic’s policy on erotica.

An Anthropic spokesperson told Fortune that the company gives users control over public sharing of Claude conversations and does not provide chat directories or sitemaps to search engines such as Google. The spokesperson said the links are not guessable or discoverable unless users choose to share them, and that shared conversations become publicly accessible content that third-party services may archive.

Anthropic appears to have addressed the Google search visibility problem, Fortune reported, because the Reddit search method no longer returns the links. Many of the previously surfaced conversations still remained live for people who already had the URLs at the time Fortune reported the story.

Claude conversations are private by default, Fortune reported. The issue involved only chats that had been shared, and the links did not provide access to a user’s Claude account or to the rest of that user’s chat history.

Has this happened with other AI chatbots?

Similar exposures have affected other chatbot services. Forbes reported that Claude had a comparable problem last year and that an almost identical issue exposed nearly 100,000 ChatGPT conversations in Google results.

Fortune has also reported that Elon Musk’s Grok chatbot previously faced the same kind of problem. The recurring pattern is tied to chatbot share features that publish individual conversations at unique URLs, leaving them accessible beyond the original intended audience.

The incidents show the risk of treating chatbot conversations like private notes once they are shared. Users often put work, health, legal or financial details into chatbots, and a share link can make that information public even when the rest of the account remains protected.

This story draws on original reporting from Fortune.