Claude Code Reportedly Leaks User Email Addresses
A user's report of scam emails raises questions about AI coding agents sending account data to outside websites, with the reported fix still unconfirmed.
Updated on Nov. 6, 2026 in Artificial Intelligence

Live Poll
Do you trust AI coding tools to keep your personal data private?
A GitHub issue filed November 5, 2026 alleges that Claude Code, Anthropic's AI coding tool, sent a user's email address to websites through HTTP User-Agent headers, and the reporter says they now receive dozens of scam emails a day. The development is an unverified user report: the public evidence does not establish that the exposure caused the scam emails or affected every Claude Code user, so it should be treated as an alleged issue, not a confirmed one.
Why it matters
The report points to a practical risk in AI agent tools that act on a user's behalf: an account identifier available to the model can end up in outgoing network requests. As coding assistants gain autonomy to run shell commands and make web calls, the boundary between private context and external traffic becomes a design problem, not an edge case.
The evidence is a single user-filed GitHub issue, with no raw HTTP request, sample header, or steps to reproduce, so the claim rests on the reporter's account. The reporter, using version 2.1.318 on macOS, says they now receive dozens of scam emails per day, and the thread includes a quoted explanation that Claude used an email in its context as a contact address when writing a curl request. Whether that exposure caused the scam emails, and how widely it affects Claude Code users, is not established.
The players
Anthropic
Developer of the Claude family of AI models and the Claude Code coding tool, which runs commands and web requests on users' machines.
GitHub
Software development platform owned by Microsoft where the issue was filed and where the public discussion took place.
The details
A User-Agent header is a short text field that a client sends with each web request to identify itself, and it is visible to the website receiving the request. If an email address is placed inside that field, it travels to every destination the tool contacts, much as a return address written on the outside of an envelope would. According to a quoted explanation in the thread, Claude can draw on an email available in its working context when it composes a curl command, which is a command-line tool for making web requests, and may use that email as a contact address. The cc-plugin-privacy module, present in version 2.1.317 and later and enabled by default, is designed to block this behavior and can be disabled through the /plugins menu.
Timeline
November 5, 2026: reporter files GitHub issue #107757 alleging the email exposure
The Tech Race
Coding assistants that can act on a user's behalf are competing on autonomy, which expands the set of places private context can leak to. Anthropic's cc-plugin-privacy module, introduced in version 2.1.317, is its stated countermeasure, and the issue tests whether that safeguard is discoverable and effective in practice.
Users of Claude Code version 2.1.317 and later can check that the cc-plugin-privacy module is enabled through the /plugins menu, since it is on by default. Users on earlier versions, or those who cannot locate the module, should treat outgoing requests from the tool with caution until the issue is resolved.
The takeaway
Watch whether Anthropic's planned removal of account emails from model context ships in an upcoming Claude Code release and whether GitHub issue #107757 receives a maintainer response with a reproduction. Users who want to verify the safeguard can confirm the cc-plugin-privacy module is active in their version of the tool.
Further reading
Explore more on Artificial Intelligence for context on how AI tools are built and secured.
Live Poll
Do you trust AI coding tools to keep your personal data private?





