Red teaming to probe the deployed system for exploitable: Which assessment activity is this?
Hiring specialists to adversarially attack a deployed model's safeguards and report the weaknesses is red teaming.
The question
A security team periodically hires specialists to craft adversarial prompts and inputs that try to make a deployed generative model leak data or bypass its safeguards, then reports the weaknesses found. Which assessment activity is this?
Preparing for AIGP? Take the free 5-min readiness quiz →
- Red teaming to probe the deployed system for exploitable weaknesses ✓Adversarially crafting inputs to break safeguards and reporting the failures is red teaming, a named periodic assessment activity.
- Continuous accuracy monitoring against the production baseline metricA real ongoing control, but tracking accuracy differs from adversarially attacking the system to find exploits.
- Impact assessment estimating harms to individuals before deploymentA valid pre-launch evaluation, yet it forecasts harms rather than actively attempting to break the live system.
- Documentation of incidents and open risks in the monitoring registerNecessary record-keeping, but logging incidents is passive and distinct from proactively probing for weaknesses.
The trap
Treating passive monitoring or incident logging as equivalent to actively adversarial red teaming. How to remember it
Hiring specialists to adversarially attack a deployed model's safeguards and report the weaknesses is red teaming.
How many of these would you get right?
One of 1581 AIGP questions on Certsqill. Take a free five-minute check and see your score per domain — not one number, but which section to open tonight.
Test your AIGP readiness — freeMore Understanding How to Govern AI Deployment and Use questions
- Continuous monitoring with a scheduled retraining: Which ongoing governance practice most directly addresses →
- Documenting the incidents: Which activity does this call for? →
- Forecasting secondary uses and reducing the downstream: Which activity is this? →
- All 424 Understanding How to Govern AI Deployment and Use questions →
Part of the Certsqill AIGP question bank · Understanding How to Govern AI Deployment and Use ·
Every answer, right and wrong, comes with its own explanation.
