THIS article discusses a misconfiguration issue with Claude AI that led to shared conversation links being indexed by Google, exposing sensitive user data. Despite users believing shared links were private, search engines were able to crawl and index these URLs because Anthropic failed to implement proper blocking in the robots.txt file. Some indexed shares contained critical information like wallet keys and personal identifiers.
The article emphasizes the importance of understanding how share links work and the risks of exposure, noting that this is not the first instance of such a mistake in AI platforms. Anthropic has since reached out to search engines to remove these links.