Private Chats With Claude Accidentally Indexed by Google

WHO:

TechCrunch - Anthropic, the company behind the Claude chatbot.

Fortune - Users of Claude, members of a discussion channel on Reddit.

WHAT:

CNET - Hundreds of user conversations with Anthropic’s Claude were found to have been available to anyone online, covering more than 200 conversations across at least 25 pages of search results.

Fast Company - One of the exposed artifacts included a tool supposedly comparing two treatment arms of a medical trial, which contained names, ages, and treatment dates.

Fast Company - Another artifact contained what appears to be a database of access codes to Irish residential apartment blocks.

Fast Company - Thousands of conversations between Claude and its users, alongside tools created during those chats, were available for a period of time through Google searches earlier this month.

Fast Company - One conversation involved an engineer launching “a project meant to train future Security Engineers in Web/Cloud/AI Security.”

Fast Company - Some artifacts created with Claude included a database of access codes to Irish residential apartment blocks.

TechCrunch - Claude’s “share chat” feature allows users to create links that enable anyone with the assigned URL to view a conversation or project.

Fortune - Claude’s Artifacts, interactive tools and mini apps built inside Claude, also appeared in Google search results.

Fortune - A chat labeled as “shared by Anthropic” showed Claude generating explicit content, which violates Anthropic’s policy on erotica.

Fortune - At the time of writing, many of the previously exposed chat links remained live and reachable to those already with the link.

Fortune - The same type of problem affected the company last year, exposing almost 100,000 ChatGPT conversations on Google.

TechCrunch - Claude’s “share chat” feature allows users to create links for sharing conversations.

Fortune - Private conversations with Claude appeared in Google search results, including sensitive information like cryptocurrency wallet keys and personal details.

Fortune - The exposed chats included topics such as erotica, programming, work notes, and fake book reviews.

Fortune - Anthropic stated that the issue has been fixed, as the links are no longer available via Google through the method shared on Reddit.

WHERE:

Fortune - The issue was discussed on the social media site Reddit.

WHEN:

Fortune - The problem surfaced over the weekend when users discovered the issue.

WHY:

Fast Company - Users reportedly do not understand the consequences of sharing published links, leading to potential exposure of sensitive information, as highlighted by cybersecurity analyst Rachel Tobac.

Fast Company - Users reportedly do not understand the consequences of sharing published links, leading to potential exposure of sensitive information.

Fortune - The issue arose because users had utilized Claude’s “share” feature, which makes conversations publicly accessible when shared.

Fortune - Anthropic emphasized that shared links are not guessable or discoverable unless users choose to share them.

HOW:

9to5Mac - The issue arose because the shared chat pages did not include the "noindex" tag, which is typically used to prevent indexing by search engines.

TechCrunch - The issue arose because the pages lacked a "noindex" tag, despite Anthropic's robots.txt saying not to crawl them.

Techmeme - Claude chats showed in Google and Bing search results, likely because the pages lacked a "noindex" tag, despite Anthropic's robots.txt saying not to crawl them.

Fortune - Claude creates a unique URL for shared conversations, which can be indexed by search engines if not properly secured.

Fortune - The “share” feature on AI chatbots generates a web address for the conversation, which can be published and left open to search engines.

(UPDATED July 28, 2026 19:27 UTC)

In a significant privacy breach, hundreds of user conversations with Anthropic’s Claude were discovered online, spanning over 200 discussions across at least 25 pages of search results. Among the exposed materials was a tool comparing treatment arms of a medical trial, revealing sensitive details like names and treatment dates, as well as a database of access codes to Irish residential apartment blocks. Cybersecurity analyst Rachel Tobac pointed out that users often don't grasp the risks of sharing published links, which contributed to this incident, exacerbated by the absence of a "noindex" tag on the shared chat pages that would have prevented search engine indexing.

(UPDATED July 28, 2026 09:50 UTC)

Thousands of conversations between Claude and its users, along with various tools developed during those discussions, were accessible through Google searches earlier this month. One notable conversation featured an engineer who was launching a project aimed at training future Security Engineers in Web, Cloud, and AI Security, but some artifacts created during these chats included sensitive information, like a database of access codes to Irish residential apartment blocks. The issue occurred because the pages did not have a "noindex" tag, even though Anthropic's robots.txt file instructed not to crawl them, leading to confusion among users about the risks of sharing published links.

(UPDATED July 27, 2026 23:20 UTC)

Claude has introduced a new "share chat" feature that allows users to create links for sharing conversations or projects, which has led to some of these chats appearing in Google and Bing search results. However, a chat labeled as “shared by Anthropic” has raised concerns after it was found generating explicit content, violating the company's policy on erotica, and many of these chat links remain accessible to those who have them. This issue echoes a similar problem from last year when nearly 100,000 ChatGPT conversations were exposed on Google, likely due to the absence of a "noindex" tag on the pages.

(UPDATED July 27, 2026 22:25 UTC)

Anthropic, the company behind the Claude chatbot, recently addressed a significant issue where private conversations shared through its "share chat" feature were appearing in Google search results, exposing sensitive information like cryptocurrency wallet keys and personal details. Users on Reddit discovered that these exposed chats covered a range of topics, including erotica, programming, and work notes, prompting concerns about privacy. Anthropic has since confirmed that the problem has been fixed, as the links are no longer accessible via Google through the method shared on Reddit.