Anthropic Privacy Breach: Claude Shared Chats Leaked on Google Search

A major security lapse has allowed private conversations and 'Artifacts' from Anthropic's Claude AI to be indexed by Google Search. Users who utilized the 'Share' feature found their supposedly private links were accessible to the public through simple search queries. The exposed data reportedly includes highly sensitive materials such as cryptocurrency wallet keys, personal medical information, and internal corporate strategies. This indexing error effectively removed the wall of privacy users expected when sharing links with specific individuals.

The leak occurred because the URLs generated for shared Claude chats did not include instructions to prevent search engine crawlers from indexing them. This vulnerability extended to Claude’s 'Artifacts' feature, which allows users to create code snippets, websites, and documents. Once a link was shared once, it became part of the public web. Security researchers discovered hundreds of pages of personal data that remained searchable even after the initial reports surfaced. This incident highlights a significant oversight in how AI companies manage public-facing links.

Anthropic has acknowledged the issue and is working to remove the indexed links from Google’s search results. The company stated that only chats intentionally shared by users were affected, but the lack of a 'no-index' tag was a critical failure. For users in India, where data protection laws are becoming more stringent under the DPDP Act, this breach emphasizes the risks of sharing sensitive information with AI platforms. Users are advised to delete any previously shared Claude links to prevent further exposure of their private data.