ISACA • AAIR
ISACA Advanced in AI Risk (AAIR) Practice Test
Validates expertise in managing AI-related risks across three practice areas: AI risk governance and framework integration, AI risk program management, and AI lifecycle risk management, covering AI vulnerability evaluation, impact assessment, and risk lifecycle navigation.
Exam Details
Questions
598
Duration
150 minutes
Passing Score
450/800
Difficulty
AssociateLast Updated
Feb 2026
Exam Overview
The ISACA Advanced in AI Risk (AAIR™) certification is an AI-focused IT risk management credential designed to validate advanced expertise in identifying, evaluating, and managing risks that arise from artificial intelligence adoption within organizations. It covers three core practice areas: AI Risk Governance and Framework Integration, AI Risk Program Management, and AI Life Cycle Risk Management. Together, these domains address the full spectrum of AI risk—from establishing governance structures and embedding AI risk into enterprise frameworks, to executing risk programs, evaluating AI-specific vulnerabilities, conducting impact assessments, and navigating risk throughout the AI development and deployment lifecycle.
The AAIR credential is part of ISACA's suite of Advanced AI certifications, alongside the Advanced in AI Audit (AAIA) and Advanced in AI Security Management (AAISM). Unlike these related credentials, AAIR specifically equips professionals to work cross-functionally, recommend risk responses, and guide senior management in safeguarding organizations from financial, reputational, and operational harms associated with AI integration. The certification is currently in beta, with a full launch anticipated for Q2 2026.
Who Should Take This Exam
AAIR is intended for experienced IT risk and advisory professionals who already hold a recognized risk or security certification and are seeking to extend their expertise into AI-specific risk management. Eligible professionals must hold at least one active credential from the following: CISA, CISM, CRISC, CGEIT, CDPSE (ISACA credentials), or CRMP, CRMA, CGRC, CISSP, CERP, CRCM, or PMI-RMP (global designations). Because the program does not cover foundational IT risk concepts, it is best suited for mid-to-senior-level practitioners who already operate in risk management, compliance, governance, or advisory roles and need structured knowledge to address AI's unique risk profile.
Typical candidates include IT Risk Managers, Enterprise Risk Officers, AI Governance Leads, Chief Risk Officers, and Compliance Managers working in industries where AI adoption is accelerating—such as financial services, healthcare, technology, and government. It is also relevant for consultants who advise organizations on responsible AI adoption and integration strategies.
Prerequisites
Candidates must hold at least one active qualifying credential at the time of application. Accepted ISACA credentials include CISA, CISM, CRISC, CGEIT, and CDPSE. Globally recognized designations that also qualify include CRMP, CRMA, CGRC, CISSP, CERP, CRCM, and PMI-RMP. These prerequisites are non-negotiable, as the AAIR program is explicitly designed to build on existing foundational IT risk knowledge rather than introduce it.
Beyond holding a qualifying credential, candidates should have practical professional experience working in IT risk management, AI governance, compliance, or a closely related advisory function. Familiarity with enterprise risk frameworks (such as COBIT, NIST, or ISO 31000), AI concepts including machine learning and generative AI models, and cross-functional risk communication will help candidates engage effectively with the curriculum and exam content.
Exam Format
The AAIR exam consists of scenario-based multiple-choice questions delivered in a proctored setting. The exam duration is 150 minutes. Scoring uses a scaled scoring model with a maximum score of 800 points, and the passing score is 450 out of 800—consistent with the scoring methodology used across ISACA's Advanced AI certification suite. The exact number of scored questions has not been published by ISACA as of the time of writing, as the certification is currently completing its beta phase ahead of a full Q2 2026 launch.
ISACA's Advanced AI exams are delivered online with remote proctoring available. Exam fees are estimated at approximately USD 575 for ISACA members and USD 760 for non-members, with an additional USD 50 application fee and annual maintenance fees of USD 45 (members) and USD 85 (non-members). Candidates are advised to check the official ISACA credentialing page for confirmed question counts, delivery options, and final pricing once the exam officially launches.
Skills Measured
- 1.Domain 1 – AI Risk Governance and Framework Integration: Covers the establishment and integration of AI risk governance structures within enterprise risk frameworks. Candidates are expected to understand how to align AI risk policies with organizational strategy, embed AI considerations into existing risk management frameworks (such as COBIT, NIST AI RMF, ISO 31000), and ensure board and executive-level accountability for AI risk. Exact weighting has not yet been published by ISACA.
- 2.Domain 2 – AI Risk Program Management: Focuses on designing, implementing, and managing an end-to-end AI risk program. Topics include AI risk identification and classification, risk appetite and tolerance setting for AI initiatives, cross-functional collaboration with AI development and business teams, risk communication, and reporting to stakeholders. Exact weighting has not yet been published by ISACA.
- 3.Domain 3 – AI Life Cycle Risk Management: Addresses risk at each phase of the AI lifecycle—from data acquisition and model development through deployment, monitoring, and decommissioning. Candidates must demonstrate ability to evaluate AI-specific vulnerabilities (e.g., model bias, data poisoning, adversarial attacks), conduct AI impact assessments, and apply controls appropriate to each lifecycle stage. Exact weighting has not yet been published by ISACA.
Study Tips
- Review the ISACA AI Risk Management framework materials and the NIST AI Risk Management Framework (AI RMF 1.0), as these are foundational references for the governance and program management domains.
- Leverage your existing CRISC, CISM, or equivalent credential study materials to reinforce enterprise risk concepts, then map those concepts to AI-specific scenarios covered in AAIR's three domains.
- Enroll in ISACA's official AAIR Online Review Course once it launches—this on-demand course is purpose-built for the exam and developed by the same organization that writes the test.
- Practice with ISACA's official Questions, Answers & Explanations (QA&E) database, which will offer 200+ scenario-based items. Focus on understanding the reasoning behind correct answers, not just memorizing them.
- Study real-world AI governance frameworks and case studies—including the EU AI Act, OECD AI Principles, and published AI risk incidents—to develop the scenario-based reasoning skills that ISACA's advanced exams emphasize.
- Join the ISACA community forums and local chapter discussions focused on AI risk to connect with other AAIR candidates and stay current with evolving AI risk guidance and regulatory developments.
- Once ISACA publishes the official AAIR Exam Content Outline (expected ahead of the Q2 2026 launch), use it as a checklist to identify and close knowledge gaps across all three domains before scheduling your exam.
Career Benefits
As organizations across industries accelerate AI adoption, demand for professionals who can rigorously manage AI-related risks is growing rapidly. AAIR holders are positioned for roles such as AI Risk Manager, Enterprise AI Governance Lead, Chief Risk Officer, AI Compliance Manager, and senior risk consultant specializing in responsible AI. These roles are emerging in regulated industries—including financial services, healthcare, and government—where AI governance requirements are being codified through regulations such as the EU AI Act and U.S. executive orders on AI.
Salary data for AI risk professionals in the United States ranges from approximately USD 90,000 to over USD 210,000 annually, depending on role, industry, and geography. ISACA-certified professionals have historically commanded a salary premium of 10–20% over non-certified peers, according to the Robert Half Salary Guide and Global Knowledge IT Skills and Salary Report. AAIR complements existing ISACA credentials—particularly CRISC—by adding a specialized AI risk layer that distinguishes holders in a market where general IT risk expertise is common but AI-specific risk governance skills remain scarce.
Sample Questions
Preview — answers shown5 sample questions with correct answers and explanations. Start a practice session to test yourself across all 598 questions.
1. Under NIST AI RMF MEASURE 2.11, Olympus Hiring Solutions must evaluate fairness and bias in their candidate screening AI with documented results. Testing reveals equal error rates across demographic groups but different positive predictive values, meaning candidates from Group A who are recommended are more likely to succeed than those from Group B. Which fairness metric is violated? (Select one!)
APredictive parity, which requires equal positive predictive value across groups
BIndividual fairness, which requires similar individuals receive similar predictions
CDemographic parity, which requires equal selection rates across groups
DEqualized odds, which requires equal true positive and false positive rates
Explanation
Predictive parity requires that positive predictive value be equal across demographic groups. When candidates from Group A who receive positive recommendations succeed more often than those from Group B, predictive parity is violated. This matters because it means the recommendations have different reliability for different groups. Equalized odds requires equal TPR and FPR, which the scenario states are equal. Demographic parity requires equal selection rates regardless of ground truth outcomes. Individual fairness addresses treatment of similar individuals, not group-level statistical parity. Understanding which fairness metric is violated is essential for selecting appropriate mitigation strategies.
2. Meridian Financial Services implements ISO/IEC 42001 Clause 6.2 for their credit risk AI. They must set AI objectives that ensure the management system can achieve intended outcomes, enhance desirable effects, prevent undesired effects, and achieve improvement. Which objectives BEST satisfy these requirements? (Select two!)
Multiple correct answersAReduce model training time by 30 percent to accelerate deployment cycles
BAchieve demographic parity within 5 percent across all protected groups by Q3 2026
CMaintain model accuracy above 92 percent while ensuring explainability for all adverse credit decisions
DIncrease AI system usage to 100 percent of credit applications regardless of risk category
EMinimize documentation requirements to reduce administrative overhead
Explanation
Clause 6.2 requires objectives that address intended outcomes, enhance desirable effects, prevent undesired effects, and drive improvement. Achieving demographic parity prevents undesired discrimination effects and enhances fairness outcomes. Maintaining accuracy while ensuring explainability balances performance objectives with trustworthiness requirements, addressing both desirable technical outcomes and regulatory compliance needs. Reducing training time is an efficiency goal that does not address trustworthiness, risk management, or compliance outcomes. Mandating 100 percent usage ignores risk-based decision-making and could increase risks. Minimizing documentation contradicts management system requirements and could prevent rather than achieve compliance and governance outcomes.
3. A cloud services provider discovers a serious incident in a general-purpose AI model classified as systemic risk under EU AI Act Article 55. According to the regulation, what is the required timeframe for reporting serious incidents to the European Commission? (Select one!)
AWithin 30 days following quarterly review
BWithin 24 hours of discovery
CWithin 2 weeks of reaching systemic risk threshold
DWithout undue delay after incident discovery
Explanation
Systemic risk GPAI providers must track and report serious incidents to the Commission without undue delay. The EU AI Act does not specify a fixed timeframe like 24 hours but requires prompt reporting. The 2-week timeframe applies to notifying the Commission when a model reaches or is expected to reach the systemic risk threshold, not for incident reporting. The 30-day quarterly review timeframe is not specified in the regulation for incident reporting. The without undue delay standard requires reporting as soon as reasonably possible after incident discovery and assessment, reflecting the severity and potential EU-wide impact of systemic risk model incidents.
4. An AI development organization must implement continuous integration and continuous delivery for ML systems. The current maturity level involves manual model training and deployment processes. Which MLOps maturity level describes a system with automated training pipelines and continuous training enabled? (Select one!)
ALevel 3: Full autonomous ML operations
BLevel 0: Manual processes without automation
CLevel 2: CI/CD pipeline automation with rapid deployment
DLevel 1: ML pipeline automation with continuous training
Explanation
Level 1 MLOps maturity is characterized by ML pipeline automation with continuous training enabled, which represents the progression from manual processes to automated training pipelines. Level 0 involves completely manual processes including training and deployment. Level 2 adds CI/CD pipeline automation enabling rapid testing, building, and deployment of the entire ML pipeline system. Level 3 is not a standard MLOps maturity designation in widely recognized frameworks. The organization moving from manual processes to automated training pipelines with continuous training specifically transitions from Level 0 to Level 1 maturity.
5. A healthcare AI system must explain individual treatment recommendations to clinicians. Which explainability technique is most appropriate for providing actionable insights showing minimal changes needed to alter a prediction? (Select one!)
AAttention Mechanisms to visualize which input portions the model focuses on
BLIME to create local surrogate models approximating behavior around specific instances
CCounterfactual Explanations to show minimal changes required to alter predictions
DSHAP to calculate game-theory based Shapley values for feature contributions
Explanation
Counterfactual Explanations provide actionable insights by showing the minimal changes needed to alter a prediction, directly answering what would need to change for a different outcome. This is particularly valuable for clinicians seeking actionable guidance. LIME provides local explanations but focuses on feature importance, not minimal changes for different outcomes. SHAP calculates feature contributions but does not inherently provide counterfactual scenarios. Attention Mechanisms show where the model focuses but are specific to architectures like transformers and do not provide actionable change guidance. Counterfactuals directly address the use case of showing what changes would alter the recommendation.
$7
One-time access to this exam
Full access to all 598 questions
Or $15/mo for all 201 exams
Detailed explanations
Free preview stays available