Over the weekend, a number of Claude conversations and Artifacts — the small interactive apps and documents users can create inside Anthropic’s Claude — were discoverable through Google searches after Reddit users showed that queries such as “site:claude.ai/share” returned long lists of shared items. Reports indicate some of the indexed items included medical records, internal company documents, and children’s names and phone numbers.
How the exposure happened
The issue appears to be linked to Claude’s “share chat” feature, which creates URLs that allow anyone with the link to view a conversation or project. Claude’s interface warns: “Anyone with the link can view.” That wording suggests the feature is intended for sharing with friends, colleagues or small groups rather than publishing to the entire web. By contrast, other services with link sharing, like Google Docs, do not necessarily have those documents appear in public search results.
Company response
Anthropic has emphasised that share links will only appear in search results if they have been posted somewhere search engines can crawl, such as a forum or social network. In an explainer, spokeswoman Amie Rotherham said: “We give people control over sharing their Claude conversations publicly, and in keeping with our privacy principles, we do not share chat directories or sitemaps with search engines like Google. These shareable links are not guessable or discoverable unless people choose to share them themselves. When someone shares a conversation, they are making that content publicly accessible, and like other public web content, it may be archived by third‑party services.”
Anthropic therefore placed responsibility on users who publish share links in locations accessible to search engines.
What was found and prior incidents
The issue was first flagged by a Reddit user on Saturday and was first reported by 404 Media on Monday morning. TechCrunch ran a test search on Monday afternoon using the Reddit method and found no results, suggesting the exposure had been addressed.
Before the problem was fixed, Futurism reported finding “a detailed medical report of a real patient, clinical trial results that included patient names, documents sharing the names and phone numbers of primary school‑aged children, company documents marked for internal use only, and employee reviews that included personal information about workers.” Exposed Artifacts also included code and work notes. Fortune reported at least one case where a chat labelled “shared by Anthropic” showed Claude producing erotica.
Anthropic’s usage policy explicitly forbids Claude from generating sexually explicit content. Attempts to make chatbots produce disallowed material through repeated or creatively framed prompts are a recurring pattern across major AI models. It is not yet clear from the exposed chats how the erotic content was generated, and Anthropic has not provided a specific comment to TechCrunch on that instance.
Search engines’ role
Google spokesperson Ned Adriance told TechCrunch: “Neither Google nor any other search engine controls what pages are made public on the web, and these pages were indexed across many search engines. We give site owners clear controls to decide whether pages can be crawled or indexed, and we always respect those directives.” In other words, search engines index publicly accessible pages, but site owners can control crawlability and indexing.
This is not the first time shared AI chats have been indexed. Last year, Forbes reported a similar problem in which Google estimated it had indexed just under 600 Claude conversations before those pages disappeared from search results. 404 Media also reported last year that a researcher was able to scrape about 100,000 ChatGPT conversations that had been set to share publicly.
What users can do
Users who want to review which Claude chats they have set to be public can check Settings -> Privacy -> Shared Chats.
The incident highlights that link‑based sharing features can create exposure and archiving risks if a link is posted to places search engines can crawl or is otherwise collected by third‑party services.
