Users of Anthropic’s Claude chatbot reportedly discovered over the weekend that some of their conversations, for which public links had been created, may have been discoverable on search engines. This development has prompted questions regarding user control over online shared content.
Anthropic allows Claude users to generate public links for their chat threads, enabling anyone with the link to view the conversations. However, users may not have realised that this could lead to their Claude conversations appearing in public search results.
The issue reportedly began on Sunday when Reddit users found chat logs by using a specific search query on Google or Bing. While publicly shareable links make chat logs public, users intending to share content with a small group may have inadvertently shared it with a wider audience.
As of Tuesday morning, searches that previously surfaced Claude chats on Google or Bing reportedly no longer did so. Anthropic previously stated in a similar incident that chats became visible because they were shared online or on social media, and that it uses a “robots.txt” file to prevent such issues.
Jacob Hoffman-Andrews, a senior staff technologist at the Electronic Frontier Foundation, noted that the privacy offered by “anyone with a link” sharing is “fragile.” He advised users to exercise “extreme caution” when creating shareable links or to consider alternative methods like copying and pasting text or screenshotting.