The lead
Claude Code put users' account email into the User-Agent of web requests; Anthropic says the email leaves the model's context next release
The mechanism is plain: some sites ask automated clients for a contact, and the model filled in the address it had. The issue cites an earlier report of the same behaviour from about five months ago, and one commenter says the mod still let a test command through.
A user reports that Claude Code, with the account email in its context, wrote it into the User-Agent header of curl calls and web fetches as a contact address, and blames it for dozens of scam emails a day. The issue reached 830 points on Hacker News; a built-in privacy mod shipped in v2.1.317 only asks before such calls, and a comment attributed to a Claude Code engineer says the email is being removed from context next week.
JOBYou run Claude Code daily: check the privacy mod is on in /plugins today and expect more spam on your account address until the fix ships; in an interview it is a clean example of why agent permissions need a deterministic layer, not a prompt.