CCAR-P Governance, Safety, and Risk Management Practice Question

Answer analysis

Option-by-option breakdown

For each option: why learners choose it and why it is or isn't the right answer here.

• ✗

  Enable verbose debug logging of every prompt and response for later forensic review.

  Why it's wrong here

  Verbose logging is valuable for investigation but is detective, not preventive. It records what a compromised agent did without restricting what it can do, so refunds and record edits would still occur at full scale. Forensic evidence helps after an incident, but the risk committee asked for controls that limit the damage an agent can cause while it is misbehaving.

• ✓

  Require human approval for refunds above a defined monetary threshold before the action is executed.

  Why this is correct

  Threshold-based human approval inserts a checkpoint before high-impact actions, so an agent acting on a malicious instruction cannot unilaterally move large sums. It bounds financial exposure even when the agent is fully compromised. This is a preventive control on consequence size, which is precisely what reducing blast radius requires, whereas monitoring or documentation alone would not stop the loss.

• ✗

  Schedule quarterly reviews of the agent's conversation transcripts to identify emerging misuse patterns.

  Why it's wrong here

  Periodic transcript review supports trend analysis and program improvement, but it operates on a quarterly cadence and cannot contain an active incident. Between reviews, a compromised agent could issue unlimited refunds and alter orders. As a governance rhythm it complements preventive controls, yet it does not reduce the maximum impact of a single malicious or erroneous action.

• ✗

  Publish an acceptable use policy that prohibits employees from manipulating the agent.

  Why it's wrong here

  A policy communicates expectations and supports disciplinary action, but it does not technically constrain the agent's capabilities. A malicious external user or injected instruction is unaffected by an internal policy document. Since the goal is limiting blast radius from a compromised agent, a non-enforceable policy statement cannot substitute for scoped permissions or approval gates.

• ✓

  Enforce least-privilege tool scopes so the agent can only call refund and order APIs permitted for the specific workflow, with per-action authorization checks.

  Why this is correct

  Least-privilege tool scoping constrains what a compromised agent can actually do by shrinking the set of callable actions to those needed for the workflow. Combined with per-action authorization, a manipulated prompt cannot escalate into arbitrary refunds or record edits. This directly limits blast radius, unlike controls that only observe, filter, or document behavior after the fact.