Authentic Isaca Certification Exam AAIR Questions Answers

Model poisoning attacks target the training data or model parameters to degrade performance or introduce malicious behavior. Supply chain tampering introduces compromised components at vendor or integration stages. Security by design principles require embedding defenses against these threats from the earliest design stages.

Why C is Correct: According to ISACA AAIR security by design guidance, adversarial resilience and data integrity controls address both model poisoning and supply chain tampering at their root. Adversarial resilience training prepares the model to resist maliciously crafted inputs. Data integrity controls—cryptographic signing, provenance tracking, integrity verification—detect tampering in training data and model artifacts across the supply chain. Together, these form the most comprehensive defense against both attack categories.

Why A is Wrong: Data refreshes with checksums detect post-hoc data corruption but do not build adversarial resilience into the model itself. Checksums verify file integrity but cannot prevent poisoning attacks that maintain file integrity while altering data content.

Why B is Wrong: Frequent retraining and bias monitoring address performance drift and fairness but do not specifically protect against deliberate tampering. A retrained model may still be trained on poisoned data if integrity controls are absent.

Why D is Wrong: Data tokenization protects sensitive field values from unauthorized access (a privacy control) but does not address model poisoning or supply chain tampering, which can occur without accessing or exposing the sensitive field values themselves.

Within the ISACA Advanced in AI Risk framework, program management connects risk identification, control selection, treatment, monitoring, resilience, third-party oversight, and reporting to enterprise risk objectives. A risk mitigation plan should contain concrete measures that reduce identified AI risks, such as controls for data poisoning and algorithmic bias. Strategic objectives and vendor-selection criteria are governance inputs, not the actual treatment measures for the identified risk. This makes option C, Measures to address issues such as data poisoning and algorithmic bias, the strongest answer. The other choices describe narrower technical, operational, performance, or administrative considerations and do not address the primary risk-management objective in the scenario as directly. A risk practitioner should select the response that most effectively reduces the stated exposure while preserving appropriate oversight, traceability, and alignment with organizational risk tolerance and business requirements.

Credit scoring AI systems are subject to anti-discrimination regulations that prohibit using models that produce biased outcomes affecting protected classes. When bias cannot be eliminated through technical means, continuing to operate the system creates ongoing legal violations and harm to affected individuals.

Why B is Correct: According to ISACA AAIR risk treatment guidance and legal compliance obligations, removing a biased credit-scoring system from production is the appropriate response when bias cannot be technically remediated. Continuing to operate a system known to produce discriminatory credit decisions violates anti-discrimination laws (such as the Equal Credit Opportunity Act), exposes the organization to regulatory enforcement, and causes ongoing harm to affected borrowers. Risk avoidance through system withdrawal is the appropriate treatment when the risk cannot be adequately mitigated.

Why A is Wrong: Requesting senior management risk acceptance for confirmed legal violations is inappropriate because organizations cannot accept risks involving known regulatory breaches. Senior management cannot legitimately authorize continued discriminatory lending practices.

Why C is Wrong: Sourcing a replacement system is a necessary future action but takes time to procure, validate, and deploy. In the interim, the biased system should not continue operating. Removing the system from production should precede replacement planning.

Why D is Wrong: Applying compensating controls to generate offsetting biases compounds the discriminatory problem rather than resolving it. Deliberately introducing additional bias—even in the opposite direction—creates an unpredictably biased model that does not produce fair outcomes.