Preflight.ai

Warns you before personal data, passwords, API keys or confidential info is sent to ChatGPT or Claude. Checks run on-device.

Pasted an ID number, a customer's phone number or a password into an AI chat by mistake? Preflight.ai catches it before it's sent. When you press send on ChatGPT or Claude, Preflight checks your message and any attached PDF, Word or text files. If it finds sensitive data, it pauses the send and shows you what it found. You choose what happens next: • Replace with safe data: sensitive values become clear placeholders, so the AI can still help. • Send anyway: it's your call, and Preflight never blocks you. • Go back and edit. WHAT IT CATCHES • Indian ID and bank details • Phone numbers and email addresses • Passwords, API keys and other secrets • Scanned documents and photos it can't read: it asks you to check them yourself PRIVATE BY DESIGN • Every check runs inside your browser. The free version makes no network requests: nothing you type is sent to us or anyone else. • No account, no sign-up, no analytics, no ads. • It only runs on ChatGPT and Claude. • Its local history records what type of data was caught and when, never the text. YOU STAY IN CONTROL • Pause it for an hour, or switch it off on one site. • Turn off any check you don't need. • Mark a value "Not sensitive" and it won't warn you about it again. FOR TEAMS Preflight Enterprise adds private keyword lists, an admin dashboard with full history and exports, and AI checks for business documents. Preflight is a safety net, not a guarantee. It catches the common mistakes, so still take a look before you paste.