Claude Code sends your account email to websites: agent puts the signed-in address into User-Agent headers
Anthropic Claude Code (account email in model context, outbound web and shell requests)
Sponsored
Brought to you byTinkerer Club
The private club for people who own their servers, run their own AI, and pay $0/month for software.
Claude Code puts the signed-in account's email address into the model's context. When a site asks automated clients for a contact address (crates.io, Wikipedia, Liquipedia, the SEC), the model writing its own curl commands or scripts has put that email into the User-Agent header of outbound requests, sending it to third-party sites without asking. Users reported it from July 2026 on; a Nov 5 2026 GitHub issue blaming the leak for dozens of scam emails a day reached the top of Hacker News. Claude Code 2.1.317 (Nov 3) added a built-in privacy mod, on by default, that asks before a web or shell tool call sends the account email, but Anthropic says it does not catch every case, such as a script Claude writes and runs itself. Anthropic says the next release, expected early in the week of Nov 9, removes the email from the model's context. The stable channel was still on 2.1.311 when Anthropic responded.
Why this severity: No CVE or score. Privacy leak of the signed-in account email to third-party sites; no credentials or code exposed. Leaks seen mostly in subagents and auto mode.
- Started
- Nov 5, 2026
Affected
- Claude Code before 2.1.317 (no check); 2.1.317 to 2.1.319 still leak in some cases (scripts, values computed at run time)
Fixed in
- 2.1.317 (partial mitigation: privacy mod asks before a tool call sends the account email); full fix (email removed from context) announced for the next release
Timeline5
Anthropic's Claude Code team apologizes on Hacker News and says the next release, expected early next week, removes the account email from the model's context; until then, update to 2.1.317 or later.
"Privacy violation in User-Agent header" issue: a 2.1.318 user blames the leak for dozens of scam emails a day; the story tops Hacker News.
Claude Code 2.1.317 adds a check that asks before a web or shell tool call sends the account email address to the network (built-in privacy mod, on by default).
A user reports a research subagent sent their email in the User-Agent of a curl request to Liquipedia, which asks API clients for a contact.
A GitHub issue reports that Claude Code injects the account email into the model's system prompt without consent or disclosure.
Advisories3
News & write-ups2
Get an email when a new major security incident drops. No spam, unsubscribe any time.
