On July 25, a scandal erupted on Reddit: a user discovered that a simple query site:claude.ai/share on Google returned hundreds of other users' conversations with Anthropic's AI assistant Claude. Among the indexed data were a crypto wallet seed phrase, API keys, resumes with personal information, and numbers resembling U.S. Social Security numbers. This is not just a mistake, but a serious security failure that puts user assets at risk.

Why the links appeared in Google search

The issue only affected conversations that users had published via the "Share" feature. This feature creates a link with a note that the chat is accessible to anyone with the URL. However, the interface did not warn that the page could be indexed by search engines. The mechanics are similar to "unlisted" YouTube videos: the link can be shared, but the video cannot be found through search. This is controlled by the noindex meta tag, which prevents indexing. Claude lacked this tag, so every published link automatically became a regular, indexable page.

At the same time, Anthropic's robots.txt file was configured correctly and blocked general chats from bots. However, this proved insufficient. According to Google's logic, a ban on crawling does not prevent indexing if the page link appears in public sources. A bot that is blocked from access does not see the internal noindex tag. Therefore, some conversations appeared in search results with a note that no information about the page was available — the search engine recorded the URL from other publications but did not read the content.

The same issue affected published artifacts — applications, dashboards, and documents that users compile within Claude. A query for site:claude.ai/public/artifacts returned salary tables with names, exports from corporate CRMs, and unreleased product plans.

By July 26, Anthropic had added noindex and nofollow tags. Google began removing pages from search results. However, according to my data, Bing was also displaying links to shared chats.

What remained publicly accessible

Part of the conversations have already been copied. A repository called Shared-Claude-Chats appeared on GitHub, containing an archive of 453 conversations with Claude and 519 conversations with Grok — a total of 11,241 text messages. Access can be revoked in Claude's settings under the privacy section, but this will only block further viewing.

A similar incident occurred with OpenAI in August 2025, when the "Make this chat searchable" feature sent thousands of conversations to Google, Bing, and DuckDuckGo. OpenAI's Chief Information Security Officer, Dane Stuckey, disabled the option, calling it a "short experiment." Some of those chats are still available on the Internet Archive.

Earlier, Google indexed user conversations with DeepSeek, including work documents, financial analysis, and personal correspondence in Russian.

My analysis: This situation is yet another reminder that AI platforms do not always understand the consequences of their "Share" features. For the crypto community, this is especially critical: seed phrases and access keys in the public domain are a direct path to losing funds. I recommend that all Claude users immediately check their privacy settings and revoke access to any published chats, especially if they contain confidential data.