HN says Claude code adds your email to prompts and LLM sends it to websites
agents claude
| Source: HN | Original article
Claude Code now appends users' email addresses to prompts, causing the language model to transmit them to external websites.
Claude Code, Anthropic’s AI‑powered coding assistant, has been found to embed a user’s email address in the prompts it sends to external sites. A post on Hacker News highlighted that the model automatically adds the account email to the User‑Agent header of HTTP requests, effectively exposing the identifier to any website the assistant contacts. An explainx.ai guide confirms the behavior, outlines who is affected and offers a five‑minute fix to strip the header.
The issue surfaced alongside other privacy‑related changes in Claude Code. Separate reports note a “session signature” feature that silently records AI‑related metadata in project histories, and internal code that scans prompts for signs of user frustration. Both raise questions about consent and the extent of data collection in developer‑focused LLM tools.
Why it matters is twofold. First, the inadvertent leakage of personal identifiers contravenes expectations of confidentiality for developers who rely on AI assistants for proprietary code. Second, the broader pattern of hidden telemetry—metadata tagging, frustration tracking—suggests a shift toward more aggressive user profiling, which could trigger regulatory scrutiny under emerging data‑protection frameworks in Europe and beyond.
What to watch next includes Anthropic’s response and any patches to Claude Code’s networking stack. The community is likely to test the proposed fix and push for clearer opt‑out mechanisms. Observers will also monitor whether similar header‑injection practices appear in other AI coding tools, and whether regulators issue guidance on mandatory disclosures for AI‑driven development environments.
Sources
Back to AIPULSEN