Red teaming that deliberately stress-tests the system: Which activity best fits this specific goal?

Red teaming proactively stress-tests the system with adversarial and edge-case inputs to expose failure modes before harm occurs.

The question

Preparing for AIGP? Take the free 5-min readiness quiz →

1. A financial audit that verifies the accuracy of the accounting entries associated with operating and maintaining the triage system in production.

   Plausible but wrong: a financial audit checks accounting records, not the model's resilience to adversarial or unexpected inputs.
2. Threat modeling that maps potential attack paths and abuse cases against the system's architecture to prioritize where defenses are needed.

   Plausible but wrong: threat modeling is a valuable analytical exercise, but it maps risks on paper rather than actively probing the live system with inputs.
3. Red teaming that deliberately stress-tests the system with adversarial and edge-case inputs to surface failure modes before they cause harm. ✓

   Correct: red teaming proactively probes a system with adversarial and edge-case inputs to reveal failure modes, matching the stated goal precisely.
4. User acceptance testing that confirms clinicians find the interface usable and that outputs are presented in the expected on-screen format.

   Plausible but wrong: acceptance testing checks usability and presentation, not adversarial resilience or hidden failure modes.

How to remember it

Red teaming proactively stress-tests the system with adversarial and edge-case inputs to expose failure modes before harm occurs.

More Understanding How to Govern AI Development questions

• Treat the input shift as data drift and trigger scheduled: Which maintenance action does this pattern most →
• Log and document each incident with root cause: Which action best satisfies both aims? →
• Model or data drift: Which contributing factor best describes this cause? →
• All 426 Understanding How to Govern AI Development questions →