Kavach: Indian PII shield for AI chats

Warns before you send Aadhaar, PAN, UPI, bank or phone details to ChatGPT, Claude or Gemini, and masks them. Runs on your device.

Kavach (कवच, "shield") stops your personal details from slipping into AI chats by accident. When you paste or send a message on ChatGPT, Claude or Gemini, Kavach checks it for Indian personal data. If it finds any, it pauses and shows you what kind, for example "This message contains 1 PAN and 1 phone number", with three choices: • Mask: replace them with placeholders like [PAN_1] and [PHONE_1], so the AI never sees the real values. You review the message before sending. • Send anyway: send it as it is. • Cancel: go back and edit. WHAT IT DETECTS • Aadhaar numbers (checked with the official Verhoeff check digit) • PAN • UPI IDs (name@okaxis, number@paytm and more) • IFSC codes • Bank account numbers (when words like "account", "a/c" or "khata" are nearby) • Indian mobile numbers (+91, 0 or 10-digit) • GSTIN (with its check character) • Email addresses It understands English, Hinglish and Hindi context: "mera aadhaar", "khata number", "मोबाइल". PRIVATE BY DESIGN • Runs entirely on your device. Kavach makes no network requests at all. • Your messages are never stored, logged or sent anywhere. • It only works on chatgpt.com, claude.ai and gemini.google.com, nowhere else. • It stores only on/off settings and counts like "PAN: 3", which you can reset. • Open source (MIT): https://github.com/harshita1615/kavach-pii GOOD TO KNOW • Interface in English and Hindi. • You can turn Kavach off for any site from its toolbar popup. • Kavach reduces risk but can miss things. Please check before you send. It is not legal advice and does not by itself make anyone compliant with any law. Kavach is not affiliated with OpenAI, Anthropic or Google.