Some Claude AI Chats Accidentally Became Public Through Search Engines

Hundreds of chats with Anthropic’s AI chatbot, Claude, were briefly accessible through search engines such as Google, raising fresh concerns about user privacy and the handling of shared AI conversations.

The issue came to light after users on Reddit discovered that links to shared Claude chats were appearing in Google search results. By using a simple site-specific search, people were able to find more than 200 publicly accessible conversations spread across dozens of search result pages.

The exposed chats covered a wide range of topics. Some users had asked Claude for help with resumes that included their names, contact details, and work history. Others shared confidential work-related information, including healthcare research, private meeting transcripts, and corporate project details. One conversation even included a request to draft an unpublished blog post about a cloud security project.

Some Claude AI Chats Accidentally Became Public Through Search Engines

Some chats were more unusual. In one case, a user asked Claude how to literally transform into a mythical nine-tailed fox. The chatbot initially responded by generating an image claiming the user had received “fully functional fox powers.” Another conversation showed a user asking the AI whether it wanted to help them or prioritize Anthropic’s interests, to which Claude replied that it experiences “something like wanting to help you.”

Anthropic said the issue was not caused by a data breach. According to the company, users have full control over whether they choose to share a conversation. Once a chat is shared, it becomes publicly accessible through a unique link. The company explained that such links cannot be discovered unless a user intentionally shares them.

However, privacy experts noted that Claude’s sharing feature only tells users that “anyone with the link” can view the conversation. It does not clearly warn that search engines may index those links, making them discoverable through online searches.

After the reports gained attention, Anthropic quickly prevented search engines from indexing shared chat links. As a result, the conversations no longer appear in search results. However, some of the previously indexed pages had already been archived or copied elsewhere on the internet.

See Also: Xiaomi’s MiMo AI Is Now 15x Faster Than ChatGPT and Claude, Without a Single Custom Chip

Google clarified that it does not decide which web pages become public. The company explained that website owners control whether their pages can be crawled and indexed by search engines, and Google follows those instructions. This suggests Anthropic likely updated its website settings to block search engines from indexing shared conversations.

The incident is not the first of its kind. Last year, OpenAI faced a similar situation when some shared ChatGPT conversations became searchable online. Following the incident, the company changed how publicly shared chat links worked. Elon Musk’s AI chatbot Grok also experienced a similar issue, with hundreds of thousands of shared conversations becoming searchable through the web.

The latest case serves as an important reminder for AI users to be careful when using sharing features. Conversations with chatbots may contain personal, financial, or business information, and once a chat is made public, it can potentially be indexed, archived, or shared beyond its intended audience. Before sharing AI conversations, users should review their content carefully and avoid including sensitive or confidential information.