User Data Classification and Sensitivity Levels
What You'll Learn
You'll learn how to categorize user data by sensitivity level and apply appropriate handling policies to protect privacy and comply with regulations.
Why Data Classification Matters
Not all data is equally sensitive. An email address requires different protection than a social security number. Data classification is the systematic process of organizing data into categories based on sensitivity, risk, and regulatory requirements—like sorting laundry by color before washing.
Core Sensitivity Levels
Public Data
Non-sensitive information that poses no risk if exposed (product catalogs, marketing content). Minimal protection required.
Internal Data
Business information not meant for public consumption (internal emails, non-sensitive analytics). Moderate access controls.
Confidential Data
Sensitive business or user data requiring strict protection:
- PII (Personally Identifiable Information): Names, addresses, phone numbers, email addresses
- Financial Data: Credit card numbers, bank accounts, transaction history
- Authentication Data: Passwords, security tokens, API keys
Restricted Data
Highly sensitive data with severe consequences if breached:
- PHI (Protected Health Information): Medical records, diagnoses, prescriptions (HIPAA-regulated)
- Special Category PII: Race, religion, political views, biometric data (GDPR Article 9)
- Legal/Regulatory Data: Court records, attorney-client communications
Applying Classification to AI Systems
When building AI applications, classify data at ingestion and apply policies accordingly:
- Storage: Encrypted at rest for confidential+, separate databases for restricted data
- Transmission: TLS for all levels, certificate pinning for restricted
- Logging: Redact confidential+ data from audit logs (as you learned in lesson 1508)
- Retention: Shorter windows for higher sensitivity, deletion on request
- Access: Role-based controls, principle of least privilege
Key Takeaway: Classifying user data by sensitivity level (public, internal, confidential, restricted) enables you to implement proportional security controls, meet compliance requirements, and minimize breach impact in AI systems.