Question 160
Original unofficial Palo Alto Networks NetSec-Architect scenario: A research group needs to decide between inline AI inspection, predeployment red teaming, and artifact scanning for a newly trained model. Which continuing assurance practice best prevents recurrence?
Explanation
Correct Answer
Original fictional security architecture situation: A research group needs to decide between inline AI inspection, predeployment red teaming, and artifact scanning for a newly trained model. Governing principle: These AI security functions apply to different lifecycle phases and cannot be treated as interchangeable. The most defensible design response is: Assign model scanning before deployment, red teaming during validation, and runtime control for live inference according to risk. Validation: The release record includes distinct results for supply-chain, adversarial, and live-traffic control gates. Sustained assurance: Update the lifecycle control matrix as models move from development to production
Incorrect Answers
“Validate that identity context is fresh, integrity protected, and available during control failure.” concerns the separate design topic: select ai security reference architectures, form factors, and data controls. It would not fix the described weakness: a research group needs to decide between inline AI inspection, predeployment red teaming, and artifact scanning for a newly trained model. These AI security functions apply to different lifecycle phases and cannot be treated as interchangeable.
This alternative is linked to the architecture category: differentiate prisma airs and ai access capabilities. The actual problem is that a research group needs to decide between inline AI inspection, predeployment red teaming, and artifact scanning for a newly trained model; the controlling distinction is: these AI security functions apply to different lifecycle phases and cannot be treated as interchangeable.
The proposed design “Include tool-result injection tests in release acceptance criteria.” addresses this related architecture topic: differentiate prisma airs and ai access capabilities. The actual security gap is that a research group needs to decide between inline AI inspection, predeployment red teaming, and artifact scanning for a newly trained model. These AI security functions apply to different lifecycle phases and cannot be treated as interchangeable.