IT requires logging of prompts and outputs for an internal policy assistant to support auditability. What responsible practice is illustrated?
Select an answer to reveal the explanation.
Short Explanation
If nobody logs what was asked and answered, you cannot audit mistakes later. Prompt and output trails keep the city honest. Publishing redacted summaries for municipal GenAI prompts and outputs won't show you who asked what and when, so you can't trace a specific answer for audit.
Full Explanation
IT requiring logging of prompts and outputs for an internal policy assistant to support auditability illustrates maintaining audit trails for municipal GenAI. The correct answer is maintaining audit trails for municipal GenAI prompts and outputs. When staff use GenAI to interpret ordinances, draft guidance, or answer policy questions, those exchanges can influence official actions. Without retained prompts, outputs, user identity, and timestamps, the city cannot reconstruct how advice was produced, investigate errors, respond to records requests, or demonstrate control to auditors.
Storing encrypted backups for municipal GenAI prompts and outputs fails because encryption and backup protect against loss and unauthorized disclosure but do not by themselves create the indexed, attributable trail with user identity and timestamps needed to reconstruct how advice was produced and audit a specific exchange. Publishing redacted summaries for municipal GenAI prompts and outputs fails because publishing redacted summaries for external review does not create the retained internal trail needed to audit how advice was produced, and redaction removes user identity and timestamps needed to reconstruct a specific exchange. Enforcing retention limits for municipal GenAI prompts and outputs fails as the tested practice because it controls how long logs are kept and when they are deleted, but it does not create the logs of prompts and outputs that demonstrate how policy advice was generated.
Exam caveat: logging is not unlimited surveillance. The tested practice is auditable trails for municipal GenAI used in official work; retention must still respect privacy minimization, privilege, and schedule rules—log enough to reconstruct, not every irrelevant token forever without policy.
Operational check: route the policy assistant through a city-controlled tenant, log user, prompt, output, and model version; restrict log access; align retention with records schedules; prohibit personal-account use for official policy queries; and periodically sample logs for policy violations and hallucinated guidance.
Prefer to study offline? Printable PDF study book + test engine ($7.50, unofficial). These questions stay free here.