dataguard — protect your data in AI chats

Warns you before you paste passwords, credit cards or medical data into ChatGPT, Claude, Gemini or any AI chatbot. 100% local.

AI chatbots are incredibly useful — but they're also a new vector for accidental data leaks. People routinely paste confidential information without thinking: a credit card to "check if this receipt is correct", an API key to "help debug this config", a medical history to "explain my diagnosis". Once sent, that data is processed by a third-party server you don't control. dataguard sits between your keyboard and the send button. It scans what you're about to submit and alerts you before you share something sensitive. WHAT IT DETECTS: - Credit / debit card numbers (Luhn-validated) — Critical - IBAN / bank account numbers — Critical - API keys and secret tokens (sk-, Bearer, AKIA...) — Critical - US Social Security Numbers — Critical - French Numéro de sécurité sociale — Critical - Passwords written in plain text — High - Medical / health data — High - Personal addresses — Medium HOW IT WORKS: 1. You type a message in ChatGPT, Claude, Gemini, Copilot, Perplexity or Poe. 2. When you press Enter or click Send, dataguard scans the text locally. 3. If it finds something sensitive, it shows a warning with the risk level. 4. You choose: edit the message, or send anyway. You always have the final word. PRIVACY FIRST: Everything runs 100% locally in your browser. dataguard never reads your messages after you send them, never connects to any server, and never stores anything. Free and open source (MIT).