Chatbot's Sensitive Chats Exposed in Google Search

Anthropic users may be sharing far more than they realize. Chat logs and projects whipped up with the Claude AI chatbot—including what appear to be medical reports, clinical trial data with patient names, children's contact details, internal corporate documents, and worker reviews—are showing up in Google search results, Futurism reports. The exposure stems from Claude's "Share" feature, which generates public links and notes that "anyone with the link can view"—but which doesn't explicitly warn that the same content can then be indexed by search engines. Some material was also published via Claude's "Artifacts" tool, which does flag potential visibility in online searches.

Anthropic says the system is functioning as designed and that only users who choose to share make their content public. "These shareable links are not guessable or discoverable unless people choose to share them themselves," a spokesperson says. TechCrunch notes, however, that the language on Claude's interface "clearly implies that the feature is mainly intended to allow users to share their chats with friends, colleagues, and small groups—not the whole internet. Google Docs, for example, offers a similar sharing feature and those documents don't end up publicly accessible on Google." Similar incidents involving Claude were reported last year by Forbes.