PII Masking, Audits, and Post-Market Monitoring

What This Is

This lesson brings governance into the operating loop: handling sensitive data, preserving auditability, and defining what to watch after launch.

Why This Matters in Production

Many teams think launch is the finish line. For AI products, launch is the beginning of continuous evidence collection about drift, misuse, and user harm.

Mental Model

Post-market monitoring means you assume the system will surprise you. Your job is to create the telemetry, review loop, and intervention paths needed when it does.

Deep Dive

PII handling matters because traces, prompts, and review content may capture sensitive information. Auditability matters because you need to explain which model version, prompt version, rubric, and evidence path led to a result. Monitoring matters because quality can drift with prompt changes, context changes, or user behavior changes. Operational maturity means you plan reviews, thresholds, and rollback triggers before they are needed.

Worked Example

A lesson-review system starts receiving pasted customer data inside submissions. A good design masks or redacts before sending to the model, preserves enough metadata for internal review, and alerts the team when patterns suggest policy misuse.

Common Failure Modes

Common failures include retaining everything indefinitely, logging raw sensitive text by default, and having no owner or cadence for post-launch review.

official-doc

OWASP User Privacy Protection

Useful for redaction and privacy-aware logging.

law

EU AI Act Overview

Use this as policy context, even if the product is not high-risk today.

official-doc

NIST AI RMF Govern Function

Tie monitoring and governance to a formal framework.