Hundreds of conversations posted via Anthropic’s AI chatbot Claude were briefly accessible in search engine results, revealing personal and work-related information that users might not have anticipated being searchable online.
The issue was with Claude’s “conversation-sharing” feature, which allows users to create a public link to a conversation. The issue was first reported by users on Reddit, and searches on search engines like Google with site-specific queries showed more than 200 shared conversations across at least 25 pages of results.
Some of the conversations included resumes with names, contact information and employment history. Others seemed to include unpublished blog posts, corporate project details, healthcare-related research and transcripts of private discussions. While the conversations had been intentionally shared through public links, some users might not have been aware those links could be indexed by search engines.
Anthropic said users retain the ability to opt out of sharing their conversations.
"Shared conversation links are not guessable or discoverable unless people choose to share them themselves," said a company spokesperson.
“When a user shares a conversation, they are making that content publicly available, and like any other public content on the web, it may be archived by third-party services,” the spokesperson continued.
Claude’s conversation sharing interface tells users that “anyone with the link” can view a shared conversation, but it doesn’t mention that search engines may index those links if they’re made public.
Examples in the indexed conversations ranged from the mundane, such as a request for help with a resume, to the bizarre. One user asked Claude philosophical questions about whether it cared more about helping people or about Anthropic. Another asked the chatbot how to literally become a mythical nine-tailed fox. Other chats seemed to include sensitive workplace materials, like draft blog posts and corporate cloud security projects.
Over the weekend the public searchability of the chat links was removed from search results. Anthropic hasn’t publicly explained what changes it made, but the removal suggests steps were taken to prevent further indexing of shared conversation pages. Third-party archives or caches might still hold previously indexed pages.
Google said it is up to the website owner — not search engines — to decide whether publicly accessible pages can be indexed.
“We provide site owners with clear controls to determine whether pages should be crawled or indexed, and we always honour those instructions,” a Google spokesperson told the BBC.
The incident is reminiscent of previous cases involving AI chat platforms. In 2024, after shared chats became searchable online, OpenAI changed how conversations in ChatGPT were handled. At the same time, conversations with Gork, the AI chatbot built into X, also began appearing in search results.
Other search engines, including Bing, Brave and DuckDuckGo, also said they had seen some of the shared Claude conversations before they were taken down. Those companies were reached for comment.
The incident underscores persistent concerns about how well users comprehend the mechanics of AI conversation sharing with web indexing, particularly when conversations contain sensitive personal or business information.
