A sanitation route advisor samples production traces and finds possible PII in logs. What privacy-aware observability step fits?

Redacting PII from traces and adjusting logging to prevent leakage is correct because privacy-aware observability must keep debugging useful without retaining unnecessary personal data. When a sanitation route advisor’s sampled production traces show possible PII, the right step is to redact or hash sensitive fields, tighten what is logged (prompts, tool payloads, retrieved snippets), and re-sample to confirm leakage stopped. Municipal GenAIOps can retain operational signal while meeting privacy expectations for resident data.

Encrypting trace storage at rest and standardizing error logging does not stop PII from entering logs; it only protects stored data and makes errors consistent, so the underlying leakage continues unremediated. Restricting trace dashboard access and retaining full payloads only limits who can view exposed data; it keeps unredacted PII in storage instead of stopping the leak at the source. Rotating trace export credentials and compressing archived logs improves access hygiene and saves space but leaves sensitive data in traces untouched, increasing privacy risk rather than containing it.

Exam caveat: privacy-aware tracing = detect leakage, redact/minimize, adjust retention and fields—not just encrypt stored traces, retain full payloads behind restricted access, or rotate credentials and compress logs while leaving leakage in place.

Operational check: inventory logged fields in sanitation traces, apply redaction policies for names, addresses, account IDs, and other PII, reduce prompt/tool payload verbosity where possible, re-sample production traces for residual leakage, and document the logging standard in the city’s GenAI privacy runbook.