Hundreds of private conversations with Anthropic's Claude chatbot were publicly accessible through search engines after shared links were indexed by Google and other crawlers, exposing sensitive personal and professional data.
Reddit users discovered more than 200 conversations spanning at least 25 pages of search results over the weekend. The exposed logs contained curricula vitae, healthcare transcripts, confidential corporate documents, and the names and phone numbers of minors. Anthropic spokeswoman Amie Rotherham said users control sharing and that anyone who shares a conversation makes its content publicly available. Google said it does not control what pages site owners make public.
The search availability of the chat logs was removed by Monday, with Anthropic likely employing tools to block the links from results. Similar exposure issues have previously affected OpenAI's ChatGPT and xAI's Grok chatbot.