Once a publicly shareable link to user chats are created, the chat logs are public and may become discoverable on search engines. In this way, users who are looking to share AI-generated content with a small group of people may inadvertently end up making their chats accessible to anyone on the internet.

The episode has raised broader questions about how much control users truly have over the information they share with AI chatbots, particularly once that data enters the larger web ecosystem. It also comes at a time when AI systems are being increasingly used by governments for surveillance and military purposes.

“The privacy afforded by ‘anyone with a link’-style sharing of chats and documents is fragile. If you share something with a friend, and they share it with a friend, and they post it to social media, suddenly it’s findable and readable by anyone on the web,” Jacob Hoffman-Andrews, a senior staff technologist at digital privacy advocacy group the Electronic Frontier Foundation (EFF) was quoted as saying by Guardian.

Hoffman-Andrews further called out Google and Anthropic for not warning or being clearer with users that their files could be indexed by search engines when creating shareable links to their AI chatbot conversations.

This is not the first time user conversations with AI chatbots have surfaced on Google. In September last year, Anthropic said that chats became visible via Google and Bing search because they were shared online or on social media. To address this issue, the Claude-maker said it updated its robots.txt file, which is a set of instructions to tell web crawlers and search engines what they can and cannot log.

However, page owners like Anthropic can also take additional steps to prevent web crawlers from surfacing pages in search results, like inserting a ‘noindex’ HTML tag. Many of the Claude user chats that appeared in Google search results did not have this tag, according to a report by Wired. To be sure, Anthropic has previously said that it does not share chat directories or sitemaps with search engines.

How can users protect themselves?

Users who want their AI chatbot conversations to remain private need to double check their sharing settings on the AI chatbot platforms as well as other apps. Additionally, they should be extremely careful when creating shareable links. They can choose to take alternative approaches such as copying and pasting the AI-generated text or screenshotting the chat.

If your conversations with Claude is already discoverable on Google, you can easily unpublish them by following the steps below:

-Go to Claude’s desktop app
-Click on your profile icon on the bottom-left corner and choose Settings from the menu bar.
– Navigate to the Privacy tab
-Scroll down to the Your Data section and click Manage public chats.
-From there, you can delete the public threads.