Claude AI Chats Exposed in Google Search via Share Feature

Shared Claude conversations surfaced in public search

Private conversations with Anthropic's Claude chatbot became publicly accessible through Google search results over the weekend, exposing sensitive user information including cryptocurrency wallet keys, personal addresses, and names. The incident affected only conversations where users had activated Claude's share feature, which generates a public URL intended for sending chat records to specific recipients.

The exposure came to light when Reddit users discovered that a particular search phrase would retrieve extensive lists of shared Claude conversations and Artifacts—the interactive tools users build within the platform. The leaked chats covered diverse topics from programming discussions and work notes to content that violated Anthropic's own policies, including one conversation labeled as "shared by Anthropic" that contained explicit material.

How the vulnerability works

When users click Claude's share button, the system creates a snapshot of the conversation at a unique web address. These URLs are designed for limited distribution but become publicly accessible once generated. The technical issue stems from search engines automatically indexing these links, making them discoverable through specific search queries despite users' expectations of privacy.

An Anthropic spokesperson told Fortune that the company does not share chat directories or sitemaps with search engines and that shared links "are not guessable or discoverable unless people choose to share them themselves." The company emphasized that when someone shares a conversation, they make that content publicly accessible, similar to other web content that third-party services may archive.

While Anthropic appears to have resolved the immediate search visibility issue, many previously exposed chat links remained accessible to anyone with the direct URL at the time of reporting.

Industry-wide pattern of exposure

This marks the second time Claude conversations have appeared in Google search results, according to Forbes reporting on a similar incident last year. OpenAI's ChatGPT experienced a nearly identical problem that exposed approximately 100,000 conversations, and Elon Musk's Grok chatbot has faced the same vulnerability. The recurring nature of these incidents across multiple AI platforms suggests a systemic challenge in balancing shareability with privacy protection.

The shared links do not grant access to users' full accounts or complete chat histories—only the specific conversations that were shared.

Why it matters

Users increasingly treat AI chatbots as confidential thinking spaces for sensitive topics including health issues, legal questions, and work problems. Unlike shared documents, which users typically curate before distribution, chatbot conversations often contain unfiltered thoughts and sensitive details users never intended for public consumption. The fact that three major AI companies have each encountered variations of this vulnerability indicates that the industry has not yet solved the fundamental tension between making conversations shareable and keeping them private. As AI adoption accelerates in professional and personal contexts, the risk of mass data exposure through these mechanisms grows proportionally.

These details were first reported by Fortune.