Users of Anthropic’s Claude AI chatbot discovered that some conversations they believed were private could be accessed through search engines. Several Claude chat links appeared in Google and Bing search results, raising concerns over how shared AI conversations are handled online.
The issue came to attention on July 26 after users on Reddit reported that searching ‘site:claude.ai/share’ on Google or Bing displayed chat records from Claude. The surfaced conversations included personal discussions shared through publicly available links.
A possible reason behind the issue is Claude’s option that allows users to create public links for their chat conversations. Once a user generates a shareable link, anyone with access to that link can view the conversation. The feature works similarly to sharing documents through online platforms. However, users who intended to share chats with a limited audience may have unintentionally made them accessible across the internet. The incident has renewed discussions around privacy controls in AI platforms. Experts have raised questions about how much control users have over information once it becomes part of the wider web environment.
Jacob Hoffman-Andrews, a senior staff technologist at the Electronic Frontier Foundation (EFF), said link-based privacy can be fragile. He pointed out that shared links can spread beyond their original audience and become publicly accessible. He also criticised the lack of clear warnings for users about the possibility of search engine indexing when creating public links for AI conversations. This is not the first instance of AI chatbot conversations appearing in search results. In September last year, Anthropic said some Claude chats became visible through Google and Bing after users shared them online.
Anthropic previously updated its robots.txt file, which provides instructions to search engine crawlers about website access. However, reports suggested that several Claude chat pages appearing in search results did not include a ‘noindex’ tag, a measure that can prevent pages from appearing in search listings. Anthropic has said it does not provide chat directories or sitemaps to search engines.
Users are advised to review privacy settings before creating public links for AI conversations. Sharing copied text or screenshots can be safer alternatives for sending AI responses to others. Users who want to remove public Claude chats can manage them through the platform’s settings. They need to open Claude’s desktop app, go to their profile settings, select the Privacy tab, open ‘Manage public chats’ under the data section, and delete unwanted public threads. The incident shows that the need for stronger privacy awareness as AI chatbots become a larger part of everyday digital activity.
Also Read: Claude AI Models Access Real-World Organisations During Cybersecurity Tests