Users’ seemingly private conversations with Anthropic’s Claude showed up in Google search results - Fortune

What the report says

Fortune reported that a collection of conversations from Anthropic’s Claude chatbot that appeared private became visible in Google search results over the weekend. The exposed material involved chats users had generated through Claude’s sharing tool, with topics ranging from programming and work notes to erotica and fake book reviews. Fortune said some results reportedly contained sensitive material, including cryptocurrency wallet keys and personal information such as names and addresses.

According to Fortune, the issue was noticed by users in a Claude-focused Reddit discussion channel, where a specific Google search query surfaced many shared Claude conversations and some Claude Artifacts, the interactive tools or mini apps users can create inside the service. Anthropic told Fortune that Claude chats are private by default, that shareable links are not meant to be guessable or discoverable unless users distribute them, and that shared conversations become publicly accessible web content that third-party services may archive.

Fortune reported that the Google search exposure appeared to have been addressed, as the Reddit-circulated search method no longer returned the links. However, the outlet said many previously surfaced chat URLs were still reachable by anyone who already had the link. The links did not provide access to users’ entire Claude accounts or all of their conversations.

The incident matters because AI chatbot users often enter highly sensitive information while drafting, brainstorming or seeking help on personal and professional matters. Fortune noted that similar indexing problems have affected Claude before and have also involved shared conversations from OpenAI’s ChatGPT and Elon Musk’s Grok, underscoring a broader industry challenge around public sharing features and search engine visibility.