AMAADOR ACADEMYCertificatesVerify a certificate

PRO-AIGP

Final check — AIGP - AI Governance Professional - Exam-Prep

Answer the 100 questions. You need 80% or more to pass. You can retry as many times as you like.

Study the course →

Question 1A retail company trains a model that learns purchasing patterns directly from historical transaction data without being explicitly programmed with rules. Which category of machine learning best describes a model trained on data that includes the known outcomes (e.g., 'fraud' or 'not fraud') for each example?
Question 2Which of the following best characterizes the relationship between artificial intelligence, machine learning, and deep learning?
Question 3A governance lead is mapping activities to the AI system lifecycle. During which lifecycle stage should requirements such as intended purpose, context of use, and acceptable risk thresholds primarily be defined?
Question 4The NIST AI Risk Management Framework (AI RMF 1.0) is organized around four core functions. Which set correctly lists all four?
Question 5Within the NIST AI RMF, which core function is described as cross-cutting and intended to be cultivated and present throughout the entire risk management process rather than executed as a discrete, sequential step?
Question 6ISO/IEC 42001 is best described as which type of standard?
Question 7A company wants third-party certification demonstrating that its AI governance processes meet an auditable, internationally recognized standard. Which standard is specifically structured to support certification of an AI management system?
Question 8The OECD AI Principles include a value-based principle that AI systems should be transparent and that responsible disclosure should enable people to understand and challenge AI outcomes. This principle is most closely associated with which OECD value?
Question 9A model performs extremely well on its training data but poorly on new, unseen data. Which concept best describes this problem?
Question 10Which statement most accurately distinguishes the NIST AI RMF from the OECD AI Principles?
Question 11During a timed AIGP mock exam, a candidate encounters a difficult scenario question and has already spent over three minutes on it. According to sound exam strategy, what is the most effective action?
Question 12In the AI system lifecycle, 'data collection and processing' precedes model building. From a governance perspective, why is documenting data provenance and representativeness at this stage critical?
Question 13An AIGP candidate reads a long scenario describing an HR screening tool, then a question asks which framework function would 'establish organizational culture and accountability structures.' Which NIST AI RMF function does this describe?
Question 14Which of the following is a characteristic that the NIST AI RMF identifies as part of a 'trustworthy' AI system?
Question 15A candidate must choose the BEST answer where two options seem partially correct. Which exam-strategy principle most directly applies to AIGP scenario questions?
Question 16The OECD AI Principles were adopted in 2019 and later updated. They are intended to apply to whom?
Question 17A data scientist splits a dataset into training, validation, and test sets. What is the primary purpose of keeping a separate test set untouched until the end?
Question 18Within the NIST AI RMF, an organization is cataloging the context, intended purpose, and categorizing the AI system to understand potential impacts before measurement. Which function is the organization performing?
Question 19A foundational concept question asks how a generative AI model (e.g., a large language model) fundamentally differs from a traditional discriminative classifier. Which answer is most accurate?
Question 20An organization adopts both the NIST AI RMF and ISO/IEC 42001. How do these two instruments most usefully complement each other?
Question 21A healthcare startup is training a model to predict patient deterioration. The team discovers that historical hospital data underrepresents a particular ethnic group due to past access barriers. Which responsible AI principle is most directly threatened if the team proceeds without remediation?
Question 22During design of a credit-scoring system, a governance lead insists on documenting the intended purpose, prohibited uses, and target population before any data is collected. What governance practice does this best represent?
Question 23A team scrapes a large web corpus to train a generative model. The legal team flags that personal data appears in the corpus without a lawful basis. From a data governance standpoint, what is the most appropriate first action?
Question 24An organization wants to distinguish between 'allocative harm' and 'representational harm' from its AI. Which scenario is an example of representational harm?
Question 25A data scientist proposes using a proxy variable (zip code) that strongly correlates with race for a model that cannot legally use race. What is the primary governance concern?
Question 26In the AI system lifecycle, at which stage is it most cost-effective to embed responsible AI requirements such as fairness constraints and privacy controls?
Question 27A model card for a training run should, at minimum, document the data sources, known limitations, and intended use. What is the principal value of this documentation during development?
Question 28An AI ethics committee reviews a proposed emotion-recognition feature for a children's education app. Which ethical consideration should weigh most heavily given the user population?
Question 29A team labels training data using low-cost crowd workers without clear guidelines, producing inconsistent labels. From a data governance perspective, what risk does this most directly create?
Question 30Which of the following best describes 'data provenance' as a data governance control during model development?
Question 31A generative model occasionally outputs verbatim copyrighted text found in its training data. Which development-stage risk does this best illustrate?
Question 32An organization adopts a 'privacy by design' approach for a recommendation model. Which technique most directly supports this principle during data preparation?
Question 33A vendor offers a pre-trained foundation model. Before fine-tuning it on internal data, the governance team asks for the model's training data documentation and known limitations. What responsible AI value does this request primarily serve?
Question 34During training, a team notices their model achieves 99% accuracy on training data but performs poorly on a held-out set. Beyond accuracy, what responsible AI concern does overfitting raise when deployed?
Question 35A company defines 'human-centric' AI as a core principle. Which design decision best operationalizes this principle in a content-moderation system?
Question 36A team must decide whether to include a sensitive attribute (e.g., disability status) in training to measure and mitigate bias. Which statement reflects current responsible AI thinking on this tension?
Question 37An AI risk taxonomy distinguishes 'inherent risk' from 'residual risk'. After applying bias mitigation and access controls to a hiring model, the remaining risk is best described as:
Question 38A developer wants to reduce the environmental harm of training a very large model. Which responsible AI consideration does this address, and what is a sound mitigation?
Question 39During data governance review, the team finds the training set was collected for a different purpose (customer support) than the new model's purpose (employee performance prediction). What principle is most at risk?
Question 40A governance lead is asked how 'explainability' should influence model selection during development for a high-stakes lending decision. What is the most defensible position?
Question 41A company is standing up its first enterprise AI governance program. The CISO argues governance should sit entirely within IT security; the Chief Privacy Officer argues it belongs in legal. Which approach best reflects AIGP guidance on operationalizing AI governance?
Question 42During model development, a data scientist discovers the training dataset for a hiring model contains 80% resumes from male applicants. Under sound bias-mitigation practice, what is the most appropriate FIRST step?
Question 43An organization wants to embed AI governance using a recognized management-system standard that supports certification and a Plan-Do-Check-Act lifecycle. Which standard is designed for this purpose?
Question 44A validation team must confirm that a credit-scoring model performs equitably across protected groups. They compute selection rates and find the approval rate for one group is 65% of the rate for the highest group. Which fairness concept does this directly relate to?
Question 45To operationalize accountability, an AI governance program creates a centralized inventory of all AI systems with risk tiers, owners and lifecycle status. What is the primary governance value of this inventory?
Question 46A team is preparing development documentation for a high-risk AI system. Which artifact best documents a model's intended use, performance characteristics, evaluation data and known limitations for downstream users?
Question 47A bias-mitigation strategy can be applied at different stages. Reweighting training samples to balance group representation BEFORE model training is an example of which type of intervention?
Question 48Adjusting decision thresholds separately for different groups on a model's output scores to equalize outcomes is best categorized as which bias-mitigation technique?
Question 49An AI governance committee must decide which projects undergo enhanced review. The most defensible criterion for triggering deeper governance scrutiny is based on:
Question 50During validation, a team uses a held-out test set and finds 98% training accuracy but 71% test accuracy. What does this most likely indicate, and what governance documentation should capture it?
Question 51A governance program adopts the NIST AI RMF. The team wants to assign responsibilities and foster a risk-aware culture across the organization. Which NIST AI RMF function most directly addresses governance structures, roles and accountability?
Question 52A red-teaming exercise on a generative model reveals it can be prompted to produce disallowed content. From a development-documentation perspective, the result should primarily be:
Question 53To operationalize human oversight, a governance program defines who can override or halt an AI system. This control is most effective when it is:
Question 54A validation report claims a model is 'fair' because it achieves equal accuracy across groups. A reviewer notes that the base rates and error types differ. What is the key governance lesson?
Question 55An organization is operationalizing governance and must define escalation paths. A model behaves unexpectedly in production. The MOST mature program would have predefined:
Question 56When selecting a fairness/bias metric, a team weighs 'equal opportunity' (equal true positive rates) versus 'demographic parity' (equal selection rates). The choice should be driven primarily by:
Question 57A governance program wants development teams to document datasets used to train models, including provenance, collection methods and known gaps. Which artifact is purpose-built for this?
Question 58A program defines policies but struggles with adoption: teams bypass governance reviews to ship faster. The most effective operationalization fix is to:
Question 59A team validates a model only on data from the same source as training data. From a governance standpoint, the main weakness of this validation approach is:
Question 60To make governance auditable, every significant AI decision (risk acceptance, model approval, exception) should be:
Question 61A European bank deploys an AI system to evaluate consumer creditworthiness for loan applications. Under the EU AI Act's risk-based classification, how is this system most likely categorized?
Question 62Which of the following practices is expressly PROHIBITED under Article 5 of the EU AI Act?
Question 63A company procures a high-risk AI recruitment tool from a third-party vendor and rebrands it under its own name before placing it on the EU market. Under the AI Act, what is the company's role?
Question 64Under the EU AI Act, which obligation specifically applies to DEPLOYERS (users) of high-risk AI systems rather than providers?
Question 65An organization is preparing to deploy a vendor-supplied AI model into production. As part of deployment readiness, which activity is MOST essential before go-live?
Question 66What is the primary purpose of effective human oversight measures for a high-risk AI system under the EU AI Act?
Question 67A deployer of a high-risk AI system observes behavior suggesting the system presents a risk to fundamental rights that the provider did not anticipate. Under the AI Act, what should the deployer do?
Question 68Under the EU AI Act, what is the maximum administrative fine for violations involving prohibited AI practices (Article 5)?
Question 69When managing vendor and third-party AI risk, which contractual provision is MOST important for an organization deploying an externally developed model?
Question 70A multinational deploys a general-purpose AI (GPAI) model that the AI Act classifies as posing systemic risk. Which obligation specifically applies to providers of such GPAI models?
Question 71Which of the following is an example of a 'limited-risk' AI system subject mainly to transparency obligations under the EU AI Act?
Question 72An organization is establishing deployment readiness for a high-risk AI system. Which governance element ensures issues detected after launch can be traced and addressed?
Question 73Under the EU AI Act, certain deployers of high-risk AI systems must perform a specific assessment before putting the system into use. What is it?
Question 74A deployer uses a high-risk AI system but feeds it input data over which it has control. Which obligation does the AI Act place on the deployer regarding this input data?
Question 75A hospital plans to deploy a third-party AI diagnostic tool. As part of vendor risk management, what is the BEST first step before signing the contract?
Question 76Which EU AI Act timeline element is correct regarding when prohibited AI practices become enforceable?
Question 77For effective human oversight, a high-risk AI system should be designed so that the human can do which of the following?
Question 78An organization deploys a high-risk AI system that affects individuals' access to a public benefit. Under the AI Act's transparency duties to affected persons, what must the deployer typically do?
Question 79A retailer wants to deploy emotion-recognition AI to monitor employees in the workplace. How does the EU AI Act generally treat emotion recognition in workplace and educational settings?
Question 80An enterprise integrates a third-party GPAI model into its own high-risk product. To support its own compliance, what does it MOST need from the upstream model provider?
Question 81A California-based company uses an AI tool to automatically screen job applicants. Under the California Privacy Rights Act (CPRA) and the CCPA regulations on automated decision-making technology, which obligation is the company MOST likely required to address?
Question 82Colorado's Artificial Intelligence Act (SB 205) imposes a duty of reasonable care on developers and deployers of 'high-risk' AI systems. The central goal of this duty is to protect consumers from what specific harm?
Question 83An organization deploys a credit-scoring model across both the EU and the US. The GDPR's Article 22 grants data subjects a right concerning automated decisions. How does this right MOST directly intersect with US state privacy laws like the CPRA?
Question 84Six months after deploying a fraud-detection model, a governance team notices the model's precision has steadily declined as fraud patterns evolved. This phenomenon, where the statistical relationship between inputs and the target changes over time, is BEST described as:
Question 85A deployed AI medical-triage system produces an output that contributes to a patient being misdirected, causing harm. The organization's incident response plan should, as a first priority, ensure which of the following?
Question 86Under the EU AI Act, providers of high-risk AI systems must establish a post-market monitoring system. What is the primary purpose of this requirement?
Question 87A company decides to decommission an AI hiring model that is being retired. Which action is MOST critical to a responsible decommissioning process?
Question 88The Illinois Artificial Intelligence Video Interview Act requires employers using AI to analyze video interviews to do which of the following?
Question 89New York City's Local Law 144 regulates automated employment decision tools (AEDTs). A key obligation it imposes before such a tool is used is that the employer must:
Question 90During post-deployment monitoring, an organization wants to detect when incoming production data differs significantly from the data the model was trained on. The metric or process they should track is BEST described as:
Question 91An AI governance lead is mapping US federal activity. Which statement BEST characterizes the current US federal approach to AI regulation compared to the EU?
Question 92A monitoring dashboard shows that an AI loan-approval model's approval rate for one protected group has dropped sharply relative to others since deployment, even though it was balanced at launch. This is an example of what governance concern requiring response?
Question 93Which of the following is the MOST appropriate trigger for activating an AI incident response plan in a production environment?
Question 94Under the GDPR, when an organization conducts AI processing likely to result in a high risk to individuals' rights, it must perform which of the following before processing?
Question 95Several US states (e.g., Virginia, Colorado, Connecticut) have enacted comprehensive consumer privacy laws that include profiling provisions. A common feature relevant to AI is that they generally give consumers the right to:
Question 96An organization sets up automated alerting so that if a deployed model's key performance indicator falls below a defined threshold, the responsible team is notified. This control is a core component of:
Question 97When responsibly decommissioning an AI system, which consideration helps prevent the loss of organizational knowledge and supports future audits?
Question 98A deployer of a high-risk AI system under Colorado's AI Act discovers that the system has caused algorithmic discrimination. The law generally requires the deployer to do which of the following?
Question 99An AI governance team is designing a post-deployment review cadence. Which combination BEST reflects a mature, risk-based monitoring strategy?
Question 100A multinational deploys the same generative AI assistant in the EU and the US. The governance team must reconcile differing obligations. Which approach BEST manages this cross-jurisdictional 'patchwork' of AI and privacy laws?