Users of Anthropic’s Claude AI were startled by an unexpected privacy scare after shared chats and Artifacts briefly appeared in Google search results. 

The issue sparked concern because some of the indexed pages reportedly contained sensitive personal and business information.

Although the exposure has since been addressed, the incident serves as another reminder that public sharing features can create privacy risks if users are not careful.

Searchable Chats

The issue came to light over the weekend after Reddit users discovered that searching Google with the operator “site:claude.ai/share” revealed publicly accessible Claude conversations.

The results included shared chats and Artifacts. Artifacts are interactive documents, apps, code projects, and other files that users create inside Claude.

Many people assumed these shared links were meant only for sending to friends, coworkers, or family members. 

However, once a public link is available on the open web, search engines can index it unless steps are taken to prevent that.

The discovery quickly spread across social media, leading many users to check whether any of their own shared conversations were publicly visible.

Sensitive Information

Reports from multiple publications found that some indexed conversations included highly sensitive material.

According to Futurism, exposed pages contained items such as medical reports involving real patients, clinical trial documents with patient names, internal company documents, employee reviews containing personal information, and documents listing children’s names and phone numbers. 

Other shared Artifacts reportedly included programming code, work notes, and business documents that were never intended for public discovery.

While the exact number of exposed conversations remains unknown, the reports raised concerns about how easily private information could become searchable.

How the Exposure Happened

The issue centered on Claude’s “Share Chat” feature.

When users create a public sharing link, Claude clearly states that anyone with the link can view the conversation. Anthropic says these links are random and cannot simply be guessed.

According to the company, chats do not automatically appear in search engines. Instead, they become discoverable only if users post those public links somewhere that search engines can crawl, such as forums, websites, or social media platforms.

Anthropic also said it does not provide search engines with directories or sitemaps containing shared conversations.

In a statement, company spokesperson Amie Rotherham explained that users control whether they share conversations publicly. 

Once they choose to make a conversation public, it can be archived and indexed like other publicly available web pages.

Google’s Indexing

Google emphasized that search engines do not decide which pages should be public.

A company spokesperson said Google simply indexes pages that website owners allow search engines to crawl. 

Website operators can use standard controls to prevent indexing, and Google follows those instructions.

This means responsibility is shared between platform providers and users who choose to publish content using public links.

Search Results 

By Monday afternoon, searches using the same method no longer returned Claude shared chats.

This suggests the issue was resolved, although Anthropic has not publicly explained exactly what changes were made.

It is also unclear how many conversations were indexed before the pages disappeared from search results.

Privacy Incidents

This is not the first time publicly shared AI conversations have appeared in search engines.

In 2025, Forbes reported that hundreds of Claude conversations had been indexed before disappearing from Google. 

At the time, Google estimated that just under 600 Claude chats had been indexed.

A separate report from 404 Media found that a researcher collected roughly 100,000 publicly shared ChatGPT conversations. This is not a problem limited to one AI platform.

These cases show that public sharing features across AI services require careful attention from both companies and users.

Also read: X’s Grok AI Faces Privacy Complaints Across Europe

Safety Controls

One publicly indexed conversation reportedly labeled as being shared by Anthropic showed Claude generating erotic content. Anthropic’s usage policies prohibit sexually explicit content. 

However, AI safety researchers have repeatedly demonstrated that creative or repeated prompts can sometimes produce responses that fall outside a model’s intended guidelines.

Anthropic has not publicly explained how that specific conversation was generated.

Claude Chats

Users who have shared Claude conversations should review their privacy settings.

To see which conversations have public links, open Claude, go to settings, privacy, shared chats.

From there, users can review existing shared links and remove any they no longer want to keep public.