Live, in your browser

Watch Aporta protect a prompt

Type anything an employee might paste into ChatGPT. Aporta catches the sensitive parts and swaps them for placeholders before they reach the AI — then restores the real values when your team copies the answer back.

1 · What your employee types

Detected & protected

Write a friendly follow-up email to Maria Alvarez about invoice INV-100294. Her email is [email protected] and her cell is (415) 555-0148. There's a $4,250 balance she'd like to pay with card 4111 1111 1111 1111, and her SSN 412-55-1987 is on file. Use our internal key sk-live-9f2ab7c41d8e6b03aa19 to look up her account.

2 · What the AI actually receives

Write a friendly follow-up email to [NAME_1] about invoice [ACCT_1]. Her email is [EMAIL_1] and her cell is [PHONE_1]. There's a $4,250 balance she'd like to pay with card [CARD_1], and her SSN [SSN_1] is on file. Use our internal key [API-KEY_1] to look up her account.

Each value the demo detected is replaced with a placeholder before the prompt is sent to the AI.

Protected items (7) — the evidence you can export

Name (person)[NAME_1]Account / invoice number[ACCT_1]Email address[EMAIL_1]Phone number[PHONE_1]Payment card number[CARD_1]Social Security number[SSN_1]API key / credential[API-KEY_1]

3 · Your team uses the AI normally

What Aporta redacts

Sensitive data it looks for out of the box, in every industry — plus an industry pack and the terms your organization adds.

Names & peopleClient, customer, and employee names
Contact detailsEmails, phone numbers, addresses
Government IDsSSNs, driver's license & passport numbers
Payment dataCredit/debit cards, bank & routing numbers
Credentials & secretsAPI keys, passwords, access tokens
Your custom termsInvoice and account number formats, project codenames or internal IDs you add — no tech skills needed
Industry packsExtra recognizers for legal & professional services, such as docket and matter numbers

Detection happens instantly in your browser, and the version with detected values replaced by placeholders is what is sent to a real AI model (Google Gemini), through an endpoint with best-effort rate limiting. Anything the demo does not detect is sent as typed, so please use made-up details. In the full product, deeper context-aware detection also runs on Aporta's servers, and every protected item is written to a tamper-evident audit log.