Claude Code leaked emails via User-Agent; 2.1.317 adds fix

The reports trace back to July 17, 2026, when GitHub user lgualtieri75 complained that an AI agent used a real email address as the User-Agent string without any confirmation. A second user published execution logs on August 11 showing curl calls to external APIs with the email embedded, and counted five such operations in a one-hour session. Engineer Chris Lakin said his own check showed more than 10 transmissions in a month, including to the US Securities and Exchange Commission.

The behavior appears to have had a plausible origin. crates.io, the Rust package site, recommends that automated programs include contact details in their User-Agent, and the August report suggested Claude Code pulled the address from Git settings to follow that convention. A separate July 25 report said the account email was also being placed into the model's system prompt, meaning the AI could use it even if the user never typed it.

A Claude Code developer responded on Hacker News on November 5, saying that after the issue drew attention on Reddit, version 2.1.317 added the cc-plugin-privacy mechanism to stop improper transmission of personal information. It is on by default and can be turned off via /plugins, and contributors asked users to report any cases the new protection does not cover. taylor-edwards, however, argued that a setting is not enough and that the transmission mechanism itself should be removed. The link between the exposure and a reported rise in scam email has not been confirmed.