Anthropic's AI Chat Data Found Publicly Searchable on Google

An unconfirmed number of shared conversations and "Artifacts" from Anthropic's AI service, Claude, have been discovered to be publicly searchable on Google. The issue came to light over the weekend when Reddit users found that specific search queries surfaced a long list of shared conversations, reportedly including sensitive information such as health records and internal company documents.

The problem appears to stem from Claude's "Share Chat" feature. This functionality allows users to generate links that grant viewing access to a conversation or project to anyone with the provided URL. While Claude's interface warns that "anyone with the link can view," users seem to have interpreted this as intended for sharing with a limited group, not the entire internet. Similar features on platforms like Google Docs typically do not index shared documents publicly.

AI company Anthropic has commented on the matter, stating that shared links appear in search results only when they have been posted on platforms accessible to search engines, such as forums or social media posts. According to the company, links sent privately to individuals do not end up in search results. This suggests that some responsibility for the data exposure may lie with users who have publicly shared links without fully considering their potential for indexing.

The incident highlights security risks associated with using AI services and the need for clearer user guidelines on how shareable links function and how sensitive information can be protected. The full extent of the exposure, in terms of the number of users and data affected, remains unclear, and Anthropic has not provided specific details on the problem's scope or the corrective actions taken.