Hundreds of conversations between users and Anthropic's Claude chatbot were found accessible to anyone via Google and other search engines, after links people had chosen to "share" ended up being indexed publicly. Some of these shared chats contained personal and work-related information, raising privacy concerns for a chatbot used by millions and marketed as a rival to ChatGPT and Gemini. Anthropic has since removed the chats from search results, though many had already been saved and circulated online before the issue was fixed.

Reddit users first spotted the exposed conversations, which spanned more than 200 chats across at least 25 pages of search results, some from as recently as weeks ago. Examples included a user asking Claude about its loyalties, a request to draft an unpublished blog post containing corporate project details, and people sharing CVs with names and contact details, alongside apparent proprietary healthcare research. Anthropic said links were "not guessable" unless users chose to share them, but the in-app share feature did not clearly warn that shared links could appear in search results; a similar incident affected OpenAI's ChatGPT last year, prompting changes to its sharing system.