On July 25, a serious privacy vulnerability was discovered on Reddit in Anthropic's AI assistant Claude. A user found that the search query site:claude.ai/share displayed hundreds of other users' conversations with the bot, indexed by Google. Among them were critically important data: a crypto wallet seed phrase, API keys, resumes with contact information, and numbers resembling U.S. Social Security numbers.
The issue only affected those conversations that users had published via the "Share" feature. It creates a link accessible to anyone with the address, but the interface did not warn that the page could be indexed by search engines. Unlike "hidden" videos on YouTube, which use the noindex meta tag, Claude lacked this. This meant that any published link automatically became a regular indexable page.
Although Anthropic's robots.txt file was correctly configured and blocked general chats from bots, this proved insufficient. Google indexes pages if there are links to them from open sources, even if access to the content is blocked. As a result, some conversations appeared in search results with a note indicating missing information — the search engine recorded the address but could not read the content.
The same fate befell published artifacts — apps, dashboards, and documents assembled within Claude. A search for site:claude.ai/public/artifacts turned up salary tables with names, exports from corporate CRMs, and unreleased product plans. By July 26, Anthropic added noindex and nofollow tags, and Google began removing pages from search results. However, some data had already spread through copies — a repository appeared on GitHub containing an archive of 453 conversations with Claude and 519 conversations with Grok, totaling 11,241 text messages.
This is not the first such incident. In August 2025, OpenAI faced a similar problem when the "Make this chat searchable" feature sent thousands of conversations to Google, Bing, and DuckDuckGo. Previously, Google indexed users' conversations with DeepSeek, including work documents and personal correspondence in Russian.
My expert commentary: This incident is yet another reminder that AI platforms cannot yet guarantee complete data privacy. Users should be extremely cautious when using publishing features, especially when it comes to crypto wallets or corporate information. The market expects stricter security standards; otherwise, trust in such services will be undermined.