Today, major search engines including Google, Bing, and DuckDuckGo removed a significant leak of private data from Claude AI users. However, a similar leak containing Perplexity users’ files still remains accessible through these same search platforms.
Perplexity, valued at $20 billion and boasting tens of millions of active users, appears to have unintentionally exposed dozens of full customer files online. These files were easily accessible by typing standard URL patterns into the search engines.
The earlier Claude leak resulted from shareable links that didn’t warn users their content could be indexed. Because Claude failed to set proper no-index tags, search engines publicly listed these sensitive files until this morning’s cleanup.
Perplexity’s situation differs, as its indexed files are live on its domain rather than cached pages. Unlike Claude’s controlled warnings, Perplexity only notifies that “Anyone with the link can view,” missing the key fact that links might be discovered by strangers online.
Google notes that it may not show every Perplexity indexed page, meaning more specific searches could pull up additional personal files. The mishandling highlights how obscure URLs become unintended gateways to sensitive user data.
A Reddit post from July 25 first exposed indexed Claude conversations, some reportedly containing personal and corporate details. While this leak has since been addressed, the Perplexity exposure persists.
