import React from "react"; export function PiiTable() { const pii = [ { type: "Basic data", examples: [ 'Names', 'Email addresses', 'User names' ], subjects: "All users as uploaded by Customer." }, { type: "Login data", examples: [ 'Randomly generated ID', 'Passwords', 'Public keys / certificates ("FIDO2", "U2F", "x509", ...)', 'User names or identifiers of external login providers', 'Phone numbers', ], subjects: "All users as uploaded and feature use by Customer." }, { type: "Profile data", examples: [ 'Profile pictures', 'Gender', 'Languages', 'Nicknames or Display names', 'Phone numbers', 'Metadata' ], subjects: "All users as uploaded by Customer" }, { type: "Communication data", examples: [ 'Emails', 'Chats', 'Call metadata', 'Call recording and transcripts', 'Form submissions', ], subjects: "Customers and users who communicate with us directly (e.g. support, chat)." }, { type: "Payment data", examples: [ 'Billing address', 'Payment information', 'Customer number', 'Support Customer history', 'Credit rating information', ], subjects: "Customers who use services that require payment. Credit rating information: Only customers who pay by invoice." }, { type: "Analytics data", examples: [ 'Usage metrics', 'User behavior', 'User journeys (eg, Milestones)', 'Telemetry data', 'Client-side anonymized session replay', ], subjects: "Customers who use our services." }, { type: "Usage metadata", examples: [ 'User agent', 'IP addresses', 'Operating system', 'Time and date', 'URL', 'Referrer URL', 'Accepted Language', ], subjects: "All users" }, ] return ( { pii.map((row) => ( )) }
Type of personal data Examples Affected data subjects
{row.type}
    {row.examples.map((example) => (
  • {example}
  • ))}
{row.subjects}
); }