Over the weekend, numerous Claude chats and interactive mini apps—termed Artifacts—became visible in Google search results. This revelation came after Reddit users employed specific search techniques that uncovered a vast range of shared conversations. Some of these conversations reportedly included sensitive information, such as health records and personal details about children.
The problem seems to stem from Claude’s “share chat” functionality, which permits users to create links for sharing conversations publicly. The platform explicitly warns that “Anyone with the link can view,” suggesting the feature was designed primarily for sharing among small groups, not for broad public access. In contrast, Google Docs has a similar capability but does not allow documents to be searchable online unless explicitly configured to do so.
Anthropic, the company behind Claude, has pointed to user actions as the cause of the issue. In response to inquiries, they clarified that shared links only appear in search results if posted on publicly visible platforms, such as social media, and that links shared privately remain undiscoverable by search engines.
A spokesperson for the company mentioned that users are granted control over the sharing settings of their Claude conversations. In line with their privacy policies, they do not share directories or maps with search engines, insisting that these links are only accessible through user sharing. When a conversation is shared, it becomes publicly available and may be archived by external services.
The issue was initially highlighted by a Reddit user on Saturday and later covered by various media outlets. By Monday afternoon, a follow-up search on Google did not yield any results, indicating that the issue may have been addressed.
Prior to the resolution, it was reported that sensitive data, including medical records and internal company documents, had been exposed. Reports also noted that some interactions included inappropriate content, which is against Anthropic’s usage guidelines.
Anthropic’s policy strictly prohibits the generation of explicit material. The emergence of such content through clever prompting techniques has been a recurring concern across multiple AI platforms. However, the specific details of how the questionable content appeared in the exposed chats remain unclear, and the company has not provided commentary on this matter.
A representative from Google emphasized that search engines do not have control over which pages are made public; instead, website owners set parameters for indexing and crawling their pages, which Google respects.
Last year, similar problems emerged when hundreds of Claude chats became indexed by search engines. Reports indicated that Google had briefly indexed almost 600 conversations before they were removed from search listings. Although it is uncertain how this incident compares in scale, various users reported locating shared conversations using the same search strategies from last year. Additionally, last year saw a researcher successfully scrape a significant number of public ChatGPT discussions.
Users can check which Claude chats are set for public sharing by navigating to the Settings menu and selecting Privacy, followed by Shared Chats.
