Shared Claude AI chats from Anthropic's chatbot appeared in public Google search results in late July, with some containing sensitive personal, medical and workplace information, according to reporting cited in the article. The issue surfaced over the weekend, with Fortune reporting on Monday 27 July 2026 that members of a Reddit discussion channel about Claude found that typing a specific search phrase into Google pulled up a long list of shared conversations. Claude's Artifacts, described as interactive tools and mini apps, also appeared in results. Some conversations reportedly included cryptocurrency wallet keys, names and addresses, with topics ranging from erotica to programming chats to work notes to fake book reviews. One chat, labelled as shared by Anthropic, showed Claude generating explicit content, which Fortune noted goes against Anthropic's policy on erotica. Futurism reported reviewing exposed material, including what appeared to be a detailed medical report of a real patient, clinical trial results with patient names, files with names and phone numbers of primary school-aged children, company documents marked for internal use only, and employee reviews with personal worker information. The exposure related to Claude's share feature, which creates a snapshot at its own web address. Fortune reported only chats users had shared with someone else were affected, and links do not give access to full accounts. An Anthropic spokesperson said the company does not share chat directories or sitemaps with search engines, and that shareable links are not guessable or discoverable unless people choose to share them