Claude Code reportedly leaks user email addresses in HTTP User-Agent headers, Cyber Security News reports

What happened

Cyber Security News reports that a GitHub user claims Anthropic’s Claude Code sent their email address to third-party websites in HTTP User-Agent headers during web searches, without their permission. Anthropic has not confirmed that the leak happened, and the report has not been independently verified.

According to Cyber Security News, the user taylor-edwards filed GitHub issue #107757 on November 5, 2026. The user says they now receive “dozens of scam emails per day.” Their setup was Claude Code version 2.1.318 on macOS, running inside tmux. The outlet says the issue has no raw HTTP request, no sample header and no steps to reproduce the behavior. The outlet also says other users reported similar experiences, but the public evidence does not show that the alleged exposure caused the scam emails or that it affects every Claude Code user.

Cyber Security News reports that the thread points to model-written commands as a possible cause. A quoted explanation in the thread describes Claude using an email address from its context as a contact address when it wrote a curl request. The outlet describes this as a possible privacy leak through routine web access. It is not a reported theft of Anthropic’s account database.

The outlet reports these other points from the issue thread:

• A contributor, poteat, said versions 2.1.317 and later include a built-in module called cc-plugin-privacy that is designed to block this behavior. The contributor said the module is enabled by default and can be disabled through /plugins. The original reporter was already on 2.1.318.
• Some commenters could not find the module. Others said harmless local tests that contained their account email and User-Agent text ran without being blocked. Those tests did not send personal data, but they raised doubts about whether the guard is active and works.
• A comment quoted a Claude Code employee as saying Anthropic planned to remove account emails from model context in an upcoming release. That statement also acknowledged gaps in the interim protection.

The issue was still open when Cyber Security News reviewed it. No CVE, fixed version or release date for the planned change is confirmed.

Am I affected?

• Possibly, if you use Claude Code with web access and your account email is in the model’s context. The thread suggests the email could end up in agent-written requests such as curl commands, according to Cyber Security News.
• The source does not confirm which versions are affected. It also does not confirm how widespread the behavior is or whether it caused the reported scam emails.
• The reporter used version 2.1.318, which the contributor says should already include cc-plugin-privacy. Running 2.1.317 or later is therefore not confirmed to protect you.

What to do now

1. Before you approve agent-written network commands such as curl, read them, and look at any User-Agent or contact fields. Cyber Security News advises teams to review these commands and verify privacy controls before allowing web access (Cyber Security News).
2. Check /plugins to see whether cc-plugin-privacy is present and enabled in your installation. Commenters report that the module may be missing or may not block test cases, so do not treat it as proof that your data cannot leave the system.
3. If you cannot review commands reliably, consider limiting Claude Code’s web access until the issue is resolved.
4. Watch for the Anthropic release that a Claude Code employee reportedly said would remove account emails from model context. A release version and date are not confirmed.

Sources

• Cyber Security News: Anthropic’s Claude Code Reportedly Leaks User Email Addresses Through HTTP User-Agent Headers