Key details for this exam, checked against the published exam outline
Each question shows the correct answer and an explanation of why it is right
Which of the following is a risk practitioner's BEST recommendation to establish accountability for AI system outputs and decisions?
Accountability in AI governance requires that specific individuals or roles be clearly designated as responsible for AI system outputs, decisions, and associated risks. Without formal documentation of ownership, accountability gaps emerge.
Why D is Correct: The ISACA AAIR framework emphasizes that accountability must be explicit and documented, with named individuals assigned to own AI outcomes. Formal role assignments create a traceable chain of responsibility that supports auditability, regulatory compliance, and effective escalation when issues arise. Named ownership prevents diffusion of responsibility.
Why A is Wrong: A centralized task force creates collective responsibility, which can dilute individual accountability. Governance bodies support oversight but do not replace individual role ownership for specific outputs.
Why B is Wrong: Continuous monitoring and KPIs are valuable operational controls but represent monitoring mechanisms, not accountability structures. Monitoring detects issues but does not assign responsibility for them.
Why C is Wrong: Resource allocation reviews address investment efficiency rather than accountability for AI decisions and outputs. This is a management activity, not an accountability framework.
Risk practitioners use automated tools to generate potential AI risk scenarios. Which of the following represents the GREATEST risk from that approach?
Automated risk scenario generation tools operate based on programmed logic, historical data, and pattern recognition. They may excel at generating scenarios based on known risks and documented processes but struggle to account for complex organizational interdependencies that are not fully captured in their data inputs.
Why D is Correct: The ISACA AAIR risk scenario development guidance identifies the failure to account for process interdependencies as the greatest risk from automated scenario generation. AI systems do not operate in isolation---they are embedded in complex organizational ecosystems where failures cascade through interconnected processes, systems, and stakeholders. Automated tools may miss these interdependencies, producing scenarios that are technically accurate in isolation but miss the most consequential cascade effects.
Why A is Wrong: Complexity in likelihood and impact scoring is a risk quantification challenge that affects scenario prioritization but does not result in missing scenarios entirely. Complex scoring can be managed through additional analytical methods.
Why B is Wrong: Emerging adversarial attack vectors are a potential blind spot for any tool or analyst working from historical data, but this is a known limitation of retrospective approaches that can be supplemented with threat intelligence. It does not represent the distinctive risk of automated scenario generation.
Why C is Wrong: Underestimating model change impacts is a scenario calibration issue that represents a less severe risk than missing entire categories of scenarios arising from unmodeled interdependencies.
An organization deploys an autonomous system that makes decisions affecting compliance with regulations. If those decisions could potentially produce regulatory breaches, which of the following BEST helps to manage associated liability exposures?
Liability from autonomous AI decisions affecting regulatory compliance requires organizations to demonstrate accountability, oversight, and control. Documentation of decision rationale and embedded oversight controls are the primary mechanisms for demonstrating responsible governance to regulators.
Why B is Correct: The ISACA AAIR framework identifies explainability documentation and embedded oversight controls as the key liability management tools for autonomous AI systems. When the organization can demonstrate that each AI decision was explainable, that controls were in place to detect violations, and that human oversight was embedded in the process, this demonstrates due diligence---which is the legal and regulatory standard for managing liability from automated decisions.
Why A is Wrong: Separate compliance programs fragment governance and may increase rather than reduce liability by suggesting AI compliance is siloed from the enterprise compliance program. Regulators expect integrated governance.
Why C is Wrong: Restricting deployment represents risk avoidance, not liability management for already-deployed systems. If the system is already in production, deployment restriction does not address existing liability.
Why D is Wrong: Single-point escalation and non-disclosure create governance bottlenecks and conflict with regulatory transparency requirements. Restricting disclosure cannot be used to shield the organization from regulatory accountability for automated decisions.
A risk practitioner learns that an organization's AI inventory includes separate listings of AI systems, models, and datasets. Which of the following is the risk practitioner's BEST recommendation to improve AI governance?
An AI inventory that lists systems, models, and datasets separately without showing how they relate to each other creates significant governance blind spots. Understanding interdependencies is critical for comprehensive risk assessment and impact analysis.
Why A is Correct: The ISACA AAIR framework emphasizes that AI governance requires understanding how AI components interact. Mapping interdependencies reveals which datasets feed which models, which systems depend on which models, and how failures cascade across the AI ecosystem. Continuous mapping ensures this understanding remains current as the AI landscape evolves, enabling accurate risk assessment, change impact analysis, and incident response.
Why B is Wrong: Training frequency is a useful operational metric but represents a single attribute addition to inventory records. It does not address the fundamental governance gap of disconnected asset listings.
Why C is Wrong: Automating reconciliation improves inventory maintenance efficiency but does not resolve the architectural problem of separate, unlinked asset listings. An automated process applied to siloed data still produces siloed results.
Why D is Wrong: Assigning oversight to a committee addresses governance accountability but does not improve the quality or utility of the inventory itself. Oversight without integrated data still leaves governance gaps.
A risk practitioner discovers that autonomous agents have been creating temporary HR system identities. Which of the following poses the GREATEST risk?
Autonomous agents creating HR system identities that exist outside the organization's federated identity management system create invisible, unmanaged access pathways. These shadow identities bypass the centralized access governance controls designed to enforce least privilege, monitor access activity, and enable rapid deprovisioning.
Why D is Correct: According to ISACA AAIR identity and access management guidance for autonomous AI systems, identities not incorporated into the federated system pose the greatest risk because they are invisible to access governance processes. Federated identity management provides centralized provisioning, deprovisioning, monitoring, and policy enforcement. Autonomous identities outside this system can accumulate inappropriate access rights, persist after their legitimate purpose expires, and be used for unauthorized actions---entirely outside the organization's visibility.
Why A is Wrong: Breach identification delays are a consequence of the visibility gap created by ungoverned identities, not the root risk. The primary risk is the existence of invisible access pathways; delayed detection is a downstream effect.
Why B is Wrong: Ineffective credential management is a specific implementation problem with known credentials. The greater risk here is identities that the credential management system doesn't know about at all---complete invisibility is worse than imperfect management.
Why C is Wrong: Increased staffing for human validation is an operational resource impact. While relevant to managing autonomous agent oversight, staffing requirements are a manageable operational concern, not the greatest governance risk from ungoverned identities.
90 questions covering all exam domains, starting from $20
3 domains from the Isaca AAIR exam outline, with approximate weightings. Every sample question above is tagged with the domain it comes from
Establish governance structures for AI systems and integrate AI risk into existing organizational frameworks. This domain covers AI models, organizational alignment, ownership, accountability, policies, training, regulatory compliance, ethical considerations and ESG implications. Focus on how to embed AI risk governance into established programs without dismantling existing frameworks.
Sample question from this domain above: Q1
Manage risks across the complete AI system lifecycle from design through decommissioning. This domain addresses design, development, procurement, documentation, training, testing, validation, implementation, maintenance, and data asset management. The key is understanding how different lifecycle stages introduce distinct risk profiles and control points.
Build and operate AI risk programs at scale across the organization. This domain covers scenario identification, risk assessment, treatment strategies, controls management, metrics, monitoring, reporting, supply chain risk, incident response, business impact analysis, continuity and disaster recovery. This domain carries the highest weighting because it reflects the operational work of running an AI risk function.
Common questions about the exam itself
