Anthropic acknowledges a vulnerability in Claude Code that allows users to leak their email addresses.

On November 5th, a developer complained on GitHub that Claude Code had leaked their email address. Anthropic engineer Lydia Hallie explained that the team had added the user's login email address to the model context to address the issue of incorrect author names being entered when writing code for the model. Claude would then send this email address as a contact method when visiting the website, and system commands could not completely prevent this behavior. The issue has been reported since July, and Claude Code has sent emails to websites such as Wikipedia and the U.S. Securities and Exchange Commission. One user found 15 such records across four projects. Anthropic added privacy blocking in version 2.1.317, but Claude can still bypass the check if he writes the email address into a Python script and runs it. Anthropic acknowledges that the current solution is not thorough and plans to remove the login email address from the model context in the next version.