AIGP Last-Week Cram Guide: What to Review the 7 Days Before the Exam

Sit a full 90-question timed mock exam under real conditions: 150 minutes, no pauses, no reference materials, no phone. Treat it as the real exam. When it's done, review every wrong answer explanation immediately — not just whether you got it right, but why the wrong options were wrong. This session is the most valuable diagnostic you can run in your final week. It identifies your specific gaps with 6 days left to address them.

After your review, categorize your wrong answers by domain. If you got 8 Domain 2 questions wrong and 2 Domain 4 questions wrong, those numbers tell you exactly where to spend Days 4 and 3. Don't let the domain breakdown be a guess — track it precisely.

The EU AI Act appears across Domain 1 (definitions), Domain 3 (risk management), and Domain 4 (regulatory landscape). In your last week, stop reading the full text of the regulation and drill the specific articles the exam tests. These are the ones that appear most frequently in AIGP exam questions:

• Article 5 — Prohibited AI Practices: Know the full list: real-time remote biometric identification in public spaces (with narrow law enforcement exceptions), social scoring by public authorities, subliminal manipulation, exploitation of vulnerabilities, and emotion recognition in workplace/educational contexts. Know what's prohibited vs. what requires safeguards.
• Article 6 + Annex III — High-Risk Classification: Two-track classification: AI systems that are safety components of products already subject to EU harmonization legislation (track 1), and AI systems in the 8 Annex III categories (track 2). Know all 8 Annex III categories.
• Article 9 — Risk Management System: Requires a continuous, iterative risk management process for high-risk AI — identification and analysis of known and foreseeable risks, estimation and evaluation, adoption of risk management measures, and residual risk assessment. Know it applies to providers of high-risk AI.
• Article 11 — Technical Documentation: Providers must prepare technical documentation before placing a high-risk AI system on the market. Know what it must contain (model description, design choices, training data info, performance metrics, risk mitigation measures).
• Article 13 — Transparency for High-Risk AI: High-risk AI systems must be sufficiently transparent that deployers can interpret the system's output and use it appropriately. The transparency obligation runs to the deployer, not the end user.
• Article 17 — Quality Management System: Providers must put in place a QMS covering design control, testing, risk management, data governance, post-market monitoring, and incident management. Must be documented and proportionate to the size of the organization.
• Article 26 — Deployer Obligations: Deployers must: use high-risk AI systems as instructed, implement human oversight measures, monitor system performance, inform providers of incidents, and (if an employer) inform workers when AI systems affecting them are deployed. Key distinction: Article 26 = deployers; Article 16 = providers.
• Article 27 — Fundamental Rights Impact Assessment (FRIA): Required for deployers who are public authorities (or bodies acting on their behalf) and for private entities deploying high-risk AI in specific Annex III contexts (credit, insurance, employment). Know who is required to perform it and what it must address.
• Article 43 — Conformity Assessment: How high-risk AI systems demonstrate compliance. Most Annex III systems self-assess (internal control procedure). Some — biometric identification, certain critical infrastructure and law enforcement uses — require third-party conformity assessment by a notified body.
• Article 73 — Serious Incident Reporting: Providers and deployers must report serious incidents to national market surveillance authorities without undue delay. The specific timeline for serious incidents (15 days for unexpected death or serious harm; 3 days where it constitutes a risk to public security) is exam-tested.
• Article 86 — Right of Explanation: Individuals affected by decisions made with the assistance of high-risk AI systems listed in Annex III have the right to obtain an explanation of the role of the AI system in the decision-making procedure and the main elements of the decision taken.

For each article: know WHO the obligation runs to (provider, deployer, or both) and WHAT specifically it requires. The exam regularly tests provider vs. deployer distinctions.

The NIST AI RMF's four core functions — Govern, Map, Measure, Manage — appear across multiple AIGP domains. Candidates who fail the AIGP often confuse these functions in scenario-based questions. In your last week, the goal is precision on what distinguishes each function.

• Govern establishes the organizational policies, accountability structures, roles, and culture that underpin all other AI risk management activities. It's not a lifecycle stage — it runs across the other three functions. Think: board-level AI policy, executive accountability assignment, culture of responsible AI.
• Map identifies and categorizes the AI risks, context, and characteristics of the specific AI system. It's discovery-phase work — who are the stakeholders, what data is used, what could go wrong, what's the intended use vs. actual use. Think: risk taxonomy, use-case documentation, impact assessment scoping.
• Measure evaluates the AI risks identified in Map using quantitative and qualitative methods. It's assessment-phase work — how severe is this risk, how likely, how does the system actually perform against defined metrics. Think: bias testing, performance benchmarking, fairness metric evaluation, red-teaming.
• Manage implements and monitors specific risk response decisions made based on Measure outputs. It's treatment-phase work — deploying risk mitigation controls, documenting residual risk acceptance, responding to incidents, updating models. Think: corrective action plans, monitoring dashboards, incident response.

Common exam confusions to drill:

• Map vs. Measure: Map discovers that a risk exists. Measure evaluates how bad it is. If a question asks about identifying which populations might be harmed → Map. If it asks about quantifying disparate impact rates → Measure.
• Govern vs. Manage: Govern creates the framework. Manage executes within it. If a question asks about establishing an AI oversight policy → Govern. If it asks about responding to a specific detected model drift → Manage.
• Govern is not the "first step" in a lifecycle sense — it's always active. It operates simultaneously with Map, Measure, and Manage.

Complete 20–25 NIST AI RMF targeted practice questions. Read every explanation — don't just tally scores.

Go back to your Day 7 mock exam domain breakdown. Which domain had the worst hit rate? Spend today entirely on that domain. If your mock exam didn't break down by domain (or you scored evenly across all), default to Domain 2 (AI Governance Program Design — 25%) and Domain 3 (AI Risk Management — 22%), since together they represent 47% of the exam. A 5-point improvement in those two domains is worth more than a 10-point improvement in Domain 5.

Format for today: read the BOK section → do 20–25 practice questions in that domain only → read every wrong-answer explanation → note the specific concept that was tested → verify your understanding of that concept from the BOK. Targeted, not broad.

Three content areas that candidates under-review and lose easy points on. They're lower volume on the exam but often appear in 2–4 questions each — questions that well-prepared candidates get right and under-prepared candidates guess on.

• Singapore Model AI Governance Framework (2nd Edition): Published by IMDA. Know the two pillars: (1) Internal Governance Structures and Measures, and (2) Operations Management. Under Pillar 1: senior management responsibility and human oversight, understanding of AI system operation and risk. Under Pillar 2: decision-making model selection, stakeholder interaction, explainability. The exam may ask which pillar a described action falls under.
• ISO/IEC 42001:2023: An AI management system (AIMS) standard — analogous to ISO 27001 for information security, ISO 9001 for quality. It certifies an organization's AI management processes, not a specific AI system or model. Published by ISO/IEC. Organizations can be certified to it by an accredited certification body. Know this in contrast to EU AI Act conformity assessment (regulatory) vs. ISO 42001 certification (voluntary management system standard).
• Domain 5 BOK Definitions (memorize the exact framing):
  • Explainability: The degree to which an explanation of how an AI system arrived at a specific output or decision can be provided in terms that are understandable to the audience.
  • Transparency: Disclosure of information about an AI system's capabilities, limitations, and use of data — to relevant stakeholders.
  • Accountability: Clear assignment of responsibility for AI systems' behavior and for addressing harms caused.
  • Human Oversight: The ability of humans to monitor, audit, intervene in, and shut down AI system operations when needed.
  • Redress: Mechanisms that allow individuals affected by AI decisions to seek review, correction, or remedy.

Stop adding content. Today is for retrieval practice only. Without looking at your notes, write down: the 5 AIGP domain names and approximate percentages. The EU AI Act risk tiers (unacceptable, high, limited, minimal). The NIST AI RMF four functions and one sentence on each. The 5 OECD AI Principles. The 8 Annex III categories from the EU AI Act. Then check each one against your notes. Any gap you find: read it once and move on. Don't drill. Don't add new content. If you're testing via OnVUE, confirm your workspace setup today: check that your ID is accessible, your room is clear, your internet is stable, and you've done the system check.

Read your cram list once — the condensed definitions and key provisions, not the full BOK. Log in 15 minutes early. Read the full question stem before looking at answer choices. Eliminate obviously wrong options first. Flag uncertain questions and return to them at the end. Do not change answers without a specific, articulable reason. Trust the preparation already done — your score is determined by the hours you've put in, not by what you read in the final hour.