Anthropic removed publicly indexed Claude conversation links from search results over the weekend after they circulated widely online. Security researchers said the move came too late to eliminate copies because some conversations had already been scraped, archived and republished elsewhere.
- Anthropic removes publicly indexed Claude conversation links from search engine results after users widely circulate shareable chat URLs online.
- Security researchers confirm that archived copies and third-party GitHub repositories continue circulating despite the original pages being deindexed.
- Similar public indexing incidents previously affected OpenAI's ChatGPT and xAI's Grok, highlighting persistent privacy vulnerabilities in AI sharing features.
The episode is the latest in a series of incidents involving major AI chatbots. Publicly shared conversations from ChatGPT, Grok and Claude have all appeared in search results over the past year. Although companies later removed or deindexed the pages, archived copies continued circulating online, underscoring how difficult it is to erase information once it enters the public web.
Claude Cleanup Follows Public Indexing
The latest incident centered on Claude’s optional chat-sharing feature, which allows users to generate public links to conversations. Researchers discovered that some of those links had been indexed by search engines, making them discoverable through ordinary searches.
Anthropic moved quickly to remove the pages from search results after the issue gained attention. Researchers said copies had already spread beyond search engines before the cleanup began.
“They deleted it. The internet didn’t,” security researcher BuBBliK wrote on X after documenting the incident.
Have a development worth tracking?
Share product launches, funding announcements, partnerships, research findings and market developments with The Grey Terminal's readership.
→ Submit a Press ReleaseA public GitHub repository created during the cleanup archived hundreds of publicly shared Claude conversations alongside previously indexed Grok chats. The repository contains conversations that had already been publicly accessible through share links rather than private user accounts.
Claude Is the Latest in a Pattern
Anthropic’s cleanup follows similar incidents involving other leading AI chatbots.
In July 2025, publicly shared ChatGPT conversations appeared in Google and other search engines after OpenAI briefly tested an option that allowed users to make shared chats discoverable. OpenAI later removed the feature after acknowledging it introduced too many opportunities for accidental exposure.
A month later, Grok generated another large-scale indexing event after hundreds of thousands of shared conversations became searchable online. Some indexed pages contained personal information, uploaded files and explicit material, prompting criticism over how public sharing was presented to users.
Claude itself experienced a similar indexing issue in September 2025, when hundreds of publicly shared conversations appeared in Google search results before being removed.
The repeated incidents have raised questions about how AI companies manage publicly shared conversations once search engines and third parties begin indexing them.
Removing Search Results Does Not Remove Copies
Security researchers said many users misunderstand what happens after search engines remove indexed pages.
Removing or deindexing a page prevents it from appearing in search results. It does not automatically delete copies already stored by archives, caches or third-party repositories.
Once a publicly shared link has been copied, downloaded or archived, those copies can continue circulating independently of the original platform.
Researchers noted that GitHub repositories, web archives and independent datasets have preserved publicly shared AI conversations after companies removed them from search results.
Privacy Depends on Sharing Settings
The incidents did not expose private AI conversations or breach chatbot systems.
Instead, the conversations involved links that users had intentionally generated through built-in sharing features. Privacy experts said users should review sharing settings before creating public links, particularly when conversations contain personal information, financial data or sensitive business material.
Most major AI platforms allow users to revoke or unshare conversations after publication. Researchers cautioned that revoking a link cannot automatically remove copies that have already been archived or downloaded elsewhere.
“The internet remembers everything you make public, even for a second,” BuBBliK wrote while documenting the latest Claude incident.
As AI assistants become increasingly integrated into work and personal life, recent incidents involving Claude, ChatGPT and Grok suggest that publicly shared conversations can outlive the search results that once pointed users to them.
Activate Terminal Layer
Structural analysis of the systems, pressures, and stakeholders behind this story.
