CAEGP Practice Quiz - Free Practice Quiz

Question 1 of 60

Which of the following BEST describes the principle of beneficence as applied to AI ethics?

A AI systems should operate without any human oversight B AI systems should minimize costs to the organization C AI systems should be profitable for their developers D AI systems should be designed and deployed to actively benefit individuals and society

1 / 60

Question 2 of 60

The OECD AI Principles include five value-based principles. Which of the following is NOT one of them?

A AI transparency and explainability B AI systems must be maximally profitable C Inclusive growth and sustainable development D Accountability for AI systems

2 / 60

Question 3 of 60

According to the IEEE Ethically Aligned Design framework, what is the PRIMARY goal of aligning AI with human values?

A To eliminate the need for human decision-making B To guarantee AI systems are always faster than humans C To ensure AI systems generate maximum economic output D To prioritize human well-being and align AI system behaviors with universally accepted human values and ethical norms

3 / 60

Question 4 of 60

Which ethical framework holds that the morality of an action is determined solely by its consequences and outcomes for overall welfare?

A Virtue ethics B Utilitarian ethics C Deontological ethics D Divine command ethics

4 / 60

Question 5 of 60

A deontological approach to AI ethics would MOST likely require which of the following?

A AI systems should be deployed only if they generate net positive economic outcomes B AI systems should emulate the virtuous character traits of their developers C AI systems should be optimized to maximize total social utility D AI systems must adhere to defined rules and duties regardless of consequences, such as always obtaining informed consent

5 / 60

Question 6 of 60

The Asilomar AI Principles were developed primarily to address which concern?

A Establishing pricing standards for AI software licenses B Optimizing AI system performance benchmarks C Standardizing AI programming languages D Ensuring long-term beneficial AI development and safety, particularly regarding advanced and transformative AI

6 / 60

Question 7 of 60

Which of the following BEST captures the principle of non-maleficence in the context of AI?

A AI systems should be transparent to their developers only B AI systems should operate without human oversight C AI systems should generate the most financial value possible D AI systems should be designed to avoid causing harm to individuals or society

7 / 60

Question 8 of 60

The principle of autonomy in AI ethics MOST directly relates to which concern?

A Ensuring AI systems can operate independently without human input B Maximizing the computational independence of AI models C Ensuring AI systems are self-correcting D Preserving individuals' rights to make informed decisions about their own lives, including the right to understand and challenge AI decisions affecting them

8 / 60

Question 9 of 60

An AI developer argues that the company's AI hiring tool should be kept entirely secret to protect intellectual property, even from candidates it rejects. Which AI ethics principle does this MOST directly conflict with?

A Explicability and transparency B Profitability C Scalability D Competitive advantage

9 / 60

Question 10 of 60

Which of the following BEST reflects the historical context of AI ethics as a formal discipline?

A AI ethics concerns have been discussed since the early days of computing, gaining significant formal structure from the 1980s onward, with major frameworks emerging in the 2010s B AI ethics emerged only after the release of ChatGPT in 2022 C AI ethics was invented by the EU as part of the GDPR legislation D AI ethics is a purely academic exercise with no practical application

10 / 60

Question 11 of 60

According to the OECD AI Principles, AI actors should be accountable for which of the following?

A Profitability of AI investments B Government AI regulation compliance only C The proper functioning of AI systems and respect for the principles throughout the AI lifecycle D Only the technical performance of AI systems

11 / 60

Question 12 of 60

Virtue ethics applied to AI governance would MOST likely focus on which of the following?

A Cultivating organizational cultures and practices where AI developers and deployers habitually demonstrate integrity, fairness, and responsibility B Maximizing efficiency of AI development processes C Minimizing regulatory compliance costs D Defining strict rules that AI systems must follow

12 / 60

Question 13 of 60

Under the EU AI Act, which of the following is classified as an UNACCEPTABLE risk AI application and is therefore PROHIBITED?

A AI used for personalized advertising B AI used for spam email filtering C Social scoring of citizens by public authorities based on personal data D AI used for inventory management

13 / 60

Question 14 of 60

Which of the following BEST describes the EU AI Act's approach to regulating AI?

A A risk-based tiered approach with different obligations for different risk levels B Voluntary guidelines with no enforcement mechanism C A blanket ban on all AI systems in the EU D Regulation only of AI systems developed within EU member states

14 / 60

Question 15 of 60

Under the EU AI Act, AI systems that pose HIGH risk are subject to which requirements BEFORE market deployment?

A Approval by the European Parliament B Only a self-declaration of conformity C Conformity assessment, registration in the EU database, technical documentation, and implementation of risk management systems D No requirements beyond standard product safety checks

15 / 60

Question 16 of 60

The NIST AI RMF's four core functions are Govern, Map, Measure, and Manage. Which function involves prioritizing and implementing risk response plans for identified AI risks?

A Map B Govern C Manage D Measure

16 / 60

Question 17 of 60

Under the NIST AI RMF, the 'Measure' function primarily involves which activity?

A Analyzing and assessing AI risks using quantitative and qualitative methods, and tracking risk over time B Deploying AI risk mitigation strategies C Establishing AI governance policies D Identifying the context and framing of AI risks

17 / 60

Question 18 of 60

ISO/IEC 42001 establishes requirements for an AI management system. Which of the following is a core requirement of this standard?

A Organizations must publish all AI training data publicly B Organizations must submit all AI systems to government certification C Organizations must establish an AI policy, define AI roles and responsibilities, and implement processes for managing AI risks and impacts D Organizations must use only open-source AI models

18 / 60

Question 19 of 60

A company establishes a cross-functional AI Ethics Board with members from legal, HR, technology, and the C-suite. This BEST exemplifies which aspect of AI governance?

A Regulatory compliance with a single AI law B Technical review of AI model accuracy C Multi-stakeholder AI governance with diverse perspectives to address the breadth of AI ethical and risk concerns D Maximizing committee size for better decisions

19 / 60

Question 20 of 60

Which of the following BEST describes AI governance maturity at the 'defined' level?

A AI governance is ad hoc and reactive B AI governance exists only in individual business units without enterprise coordination C AI governance processes are documented, standardized, and consistently applied across the organization D AI governance is fully optimized and continuously improved with data-driven feedback

20 / 60

Question 21 of 60

A national AI strategy that prioritizes domestic AI investment, workforce development, and regulatory coordination MOST directly reflects which OECD AI principle?

A AI robustness and security B AI transparency C AI accountability D Inclusive growth, sustainable development, and well-being

21 / 60

Question 22 of 60

GDPR Article 22 grants individuals specific rights regarding automated decision-making. Which of the following BEST describes the core right established?

A The right to access an AI system's source code B The right to delete all personal data from AI training datasets immediately C The right to compensation for any AI decision that is later reversed D The right not to be subject to solely automated decisions with legal or similarly significant effects, and the right to request human review

22 / 60

Question 23 of 60

Under GDPR's principle of data minimization as applied to AI systems, organizations must ensure that:

A AI systems only use data that is at least 5 years old B All personal data collected for AI is encrypted C AI systems collect and use only the personal data that is adequate, relevant, and necessary for the specified purpose D AI systems collect and process as much data as possible to improve accuracy

23 / 60

Question 24 of 60

The US Executive Order on AI Safety signed in October 2023 primarily directed federal agencies to do which of the following?

A Ban the use of AI in federal hiring processes B Establish safety testing requirements for advanced AI systems, particularly those with potential national security implications, and develop AI standards and guidance C Prohibit all use of commercial AI tools in government D Nationalize all US AI companies

24 / 60

Question 25 of 60

In the financial services sector, AI regulations primarily focus on which risk?

A Energy consumption of AI data centers B Gaming and entertainment applications of AI C Model risk in credit decisions, including fair lending obligations and explainability requirements for credit denials D Social media content generated by AI

25 / 60

Question 26 of 60

Algorithmic accountability laws, such as those enacted in some US jurisdictions, primarily require organizations to do which of the following?

A License all algorithms with the government B Use only government-approved algorithmic models C Publish the complete source code for all algorithms D Conduct and document impact assessments of automated decision systems used for consequential decisions, and provide for human review or appeals

26 / 60

Question 27 of 60

Under the EU AI Act, providers of high-risk AI systems have an obligation to register their systems in which database?

A The ISO Certified Systems Database B The EU database for high-risk AI systems (EUDB) C The EU Patent Database D The UN Global AI Registry

27 / 60

Question 28 of 60

A healthcare organization deploys an AI diagnostic tool without conducting a Data Protection Impact Assessment (DPIA) as required by GDPR. Which compliance obligation has been violated?

A EU AI Act Article 9 risk management requirement B GDPR Article 35, which requires a DPIA for processing likely to result in high risk to individuals, including systematic use of AI for health data C HIPAA audit log retention requirement D ISO 27001 certification requirement

28 / 60

Question 29 of 60

When monitoring AI systems for regulatory compliance, which of the following represents a PROACTIVE compliance approach?

A Implementing automated monitoring dashboards that continuously track AI system outputs for bias, drift, and policy violations with predefined escalation triggers B Only reviewing AI compliance following a customer complaint C Waiting for regulatory investigations to identify non-compliance D Relying solely on annual external compliance audits

29 / 60

Question 30 of 60

Which of the following sector-specific concerns is MOST relevant to AI governance in Human Resources?

A Potential violations of employment discrimination laws through biased automated screening and selection systems B The energy efficiency of AI HR systems C The cost savings from AI-automated payroll processing D The branding and marketing use of AI in HR communications

30 / 60

Question 31 of 60

Disparate impact in AI decision systems refers to which phenomenon?

A An AI system that operates faster for some users than others B An AI system that is intentionally programmed to discriminate C An AI system that treats different customer segments with different pricing D AI system outcomes that disproportionately disadvantage a protected group, even without discriminatory intent, due to biased data or model design

31 / 60

Question 32 of 60

SHAP (SHapley Additive exPlanations) is used in responsible AI implementation to achieve which objective?

A Encrypt model weights for security B Compress large AI models for edge deployment C Provide feature-level explanations showing each input variable's contribution to a specific model prediction D Accelerate model training speed

32 / 60

Question 33 of 60

Federated learning is PRIMARILY used in responsible AI to address which concern?

A Improving model accuracy by accessing more data centrally B Enabling AI model training across distributed data sources without centralizing sensitive data, thereby protecting privacy C Simplifying the AI model architecture D Reducing computational costs of AI training

33 / 60

Question 34 of 60

Differential privacy in AI systems is designed to provide which guarantee?

A The participation of any individual in a training dataset cannot be reliably inferred from the model's outputs, providing mathematical privacy guarantees B All training data is encrypted at rest C Models will produce identical outputs for all users D Models will not experience drift over time

34 / 60

Question 35 of 60

Which of the following is an example of an AI impact assessment?

A A technical performance benchmark comparing two model architectures B A systematic evaluation of the potential effects of an AI system on individuals, communities, and society, conducted before deployment C A financial analysis of AI project return on investment D An IT security penetration test of an AI system's API

35 / 60

Question 36 of 60

Human oversight mechanisms in AI governance serve which PRIMARY purpose?

A Reducing the cost of AI systems by requiring human review of all outputs B Eliminating the need for AI governance frameworks C Ensuring humans retain meaningful control and accountability over AI decisions, particularly in high-stakes contexts, and can intervene when needed D Preventing AI systems from operating without real-time human monitoring at all times

36 / 60

Question 37 of 60

LIME (Local Interpretable Model-agnostic Explanations) differs from SHAP in that it:

A Provides global model explanations rather than per-prediction explanations B Only works with neural network architectures C Approximates model behavior locally around a specific prediction by fitting a simpler interpretable model to that neighborhood, making it model-agnostic D Requires access to the model's training data to function

37 / 60

Question 38 of 60

A responsible AI development lifecycle MOST importantly includes which activity at the requirements phase?

A Selecting the AI programming framework to use B Determining the cloud infrastructure for model hosting C Writing unit tests for model functions D Defining ethical requirements, fairness criteria, and potential harms alongside functional requirements before development begins

38 / 60

Question 39 of 60

An organization's AI system is found to produce less accurate outputs for users with non-native language inputs. Which type of bias MOST likely explains this?

A Confirmation bias in human review B Representation bias in training data â€” the model was trained predominantly on native-language text, underrepresenting non-native language patterns C Algorithmic bias introduced during deployment D Recency bias

39 / 60

Question 40 of 60

Historical bias in AI datasets refers to which phenomenon?

A Bias introduced by the model's developers based on their historical preferences B Societal biases and inequalities from the past that are captured in historical data and then perpetuated or amplified by AI models trained on that data C Bias caused by using data that is more than 10 years old D Bias introduced by a model that was trained too long on historical data

40 / 60

Question 41 of 60

Measurement bias in AI systems MOST commonly arises from which source?

A Using too many variables in the model B Deploying the model in a different country than where it was trained C Using cloud computing resources for model training D Systematic errors or inconsistencies in how data was collected, measured, or labeled across different groups

41 / 60

Question 42 of 60

Aggregation bias in AI occurs when:

A A single model is applied to heterogeneous groups with meaningfully different underlying patterns, producing suboptimal or unfair results for subgroups B A model uses too many aggregated statistical summaries C Data from multiple sources is combined without deduplication D A model is trained using batch processing rather than real-time data

42 / 60

Question 43 of 60

Deployment bias in AI refers to which type of problem?

A The AI system is used in contexts or ways that differ from those for which it was designed and validated, leading to unexpected failures or harms B The model takes too long to respond in production C The deployment process did not follow the organization's change management procedures D The model was deployed using the wrong cloud provider

43 / 60

Question 44 of 60

An AI risk taxonomy typically distinguishes between which PRIMARY categories of AI risk?

A Technical risks (model failures, drift, security), operational risks (process failures, human factors), ethical/social risks (bias, privacy, autonomy), and strategic/reputational risks B Only bias risk and cybersecurity risk C Only risks specific to generative AI applications D Only technical and financial risk

44 / 60

Question 45 of 60

When communicating AI risk to a board of directors, which approach is MOST effective?

A Framing AI risks in terms of business impact, strategic implications, and required governance decisions, using plain language and visualizations B Recommending that the board take no action until a full technical audit is complete C Providing detailed technical documentation about model architectures D Sharing all model performance metrics without context

45 / 60

Question 46 of 60

Which of the following BEST describes the purpose of an AI incident reporting system?

A Documenting AI vendor contract terms B Tracking the cost of AI development projects C Reporting AI system uptime statistics to senior management D Providing a systematic mechanism to document, escalate, investigate, and learn from AI system failures, biased outputs, and unexpected behaviors

46 / 60

Question 47 of 60

A risk mitigation strategy for AI bias that involves reweighting training examples to give underrepresented groups more influence during training is BEST described as:

A In-processing algorithmic fairness constraint B Pre-processing bias mitigation through data reweighting C Deployment-phase human review D Post-hoc bias correction

47 / 60

Question 48 of 60

Under the EU AI Act, providers of limited-risk AI systems (such as chatbots) are primarily subject to which obligation?

A Prohibition from deployment B Mandatory bias audits every six months C Full conformity assessment and registration D Transparency obligations requiring users to be informed they are interacting with an AI system

48 / 60

Question 49 of 60

An AI Ethics Board is reviewing a proposed AI system for loan decisioning. The system achieves high accuracy but cannot explain individual decisions. The board should MOST likely recommend:

A Replacing the AI system with a human-only process permanently B Deferring deployment until explainability mechanisms are implemented, given the legal requirements for adverse action explanations in lending C Immediate deployment since accuracy is the primary consideration D Deploying only for low-value loans

49 / 60

Question 50 of 60

The EU AI Act's prohibited AI practices include real-time remote biometric identification in publicly accessible spaces by law enforcement. Which limited exception does the Act provide?

A Exceptions for private organizations conducting security screening B Exceptions for any law enforcement purpose at the discretion of member states C Specific, targeted exceptions for serious crimes with prior judicial authorization, subject to strict safeguards D No exceptions are permitted under any circumstances

50 / 60

Question 51 of 60

Which of the following BEST describes the relationship between the NIST AI RMF and the EU AI Act?

A They are complementary frameworks: NIST AI RMF provides voluntary risk management guidance while the EU AI Act provides mandatory regulatory requirements; organizations may use NIST to help meet EU AI Act obligations B The EU AI Act was developed directly from the NIST AI RMF C They are conflicting frameworks that cannot be used together D NIST AI RMF supersedes the EU AI Act for US-based organizations

51 / 60

Question 52 of 60

A governance professional is designing an AI ethics training program for employees. Which of the following BEST reflects the scope of training required for comprehensive responsible AI?

A Training only for data scientists and AI engineers on technical fairness techniques B Role-appropriate training covering AI ethics principles, relevant regulations, organizational AI policies, and practical guidance for each employee's interaction with AI systems C A one-time company-wide briefing on AI policy D Training only for senior executives on AI strategy

52 / 60

Question 53 of 60

Which of the following represents the GREATEST challenge to implementing explainability in deep learning models?

A The lack of available explainability tools in the market B The fundamental tension between model complexity (which drives accuracy) and interpretability (which requires simpler, more transparent structures) C Regulatory prohibition on using explainability techniques D The high cost of computing resources required

53 / 60

Question 54 of 60

An organization wants to assess whether its AI governance program meets international best practices. Which of the following assessment approaches is MOST comprehensive?

A Asking the AI development team to self-certify compliance B Checking only whether the organization has an AI ethics policy document C Conducting a structured gap analysis against multiple frameworks including NIST AI RMF, ISO/IEC 42001, and EU AI Act requirements, with evidence-based scoring D Comparing the organization's AI budget to industry benchmarks

54 / 60

Question 55 of 60

Responsible AI requires 'human-in-the-loop' oversight for high-stakes decisions. Which scenario BEST illustrates a meaningful human-in-the-loop process?

A A human is available to answer questions after decisions have already been executed B A trained reviewer examines the AI recommendation, has access to the underlying evidence and explanation, and can override or escalate the decision before it takes effect C A human presses a button to confirm AI decisions without reviewing them D A human receives AI outputs as a notification with no ability to modify the decision

55 / 60

Question 56 of 60

An organization discovers that its AI customer service system has been making systematically different warranty claim decisions for customers in certain postal codes that correlate with racial demographics. Which AI risk category does this PRIMARILY represent?

A Cybersecurity risk B Proxy discrimination bias risk â€” the model uses a facially neutral variable (postal code) as a proxy for protected characteristics C Operational resilience risk D Supply chain risk

56 / 60

Question 57 of 60

The concept of 'AI accountability' as used in the OECD AI Principles requires which of the following?

A Organizations and individuals involved in AI development and deployment must be answerable for proper functioning and respect for applicable principles throughout the AI lifecycle B AI developers must publicly disclose all model weights C AI systems must be fully autonomous without any human accountability D All AI decisions must be reviewed by an independent government body

57 / 60

Question 58 of 60

When conducting an AI impact assessment for a proposed AI recruitment tool, which stakeholder group is MOST important to consult?

A The organization's investors B The marketing department responsible for employer branding C Affected groups â€” in this case, job applicants â€” to understand potential harms and ensure the assessment captures real-world impacts on those most directly affected D The AI software vendors competing for the contract

58 / 60

Question 59 of 60

A company's responsible AI framework states that all AI systems must undergo a fairness audit before deployment. During the audit of a predictive policing tool, significant racial disparities in predicted risk scores are found. The MOST responsible course of action is:

A Deploy the system anyway since the model has high overall accuracy B Classify the finding as a minor issue and proceed with deployment C Deploy only in jurisdictions without civil rights enforcement D Halt deployment and conduct a thorough investigation of the bias sources, remediate the disparities, and only deploy after independent validation confirms equitable outcomes

59 / 60

Question 60 of 60

Which of the following BEST distinguishes AI governance from AI ethics in practical organizational terms?

A AI ethics defines the principles and values that should guide AI (the 'what' and 'why'), while AI governance provides the organizational structures, processes, policies, and accountability mechanisms that operationalize those principles (the 'how') B AI ethics is mandatory; AI governance is optional C AI governance applies only to large enterprises; AI ethics applies to all organizations D AI governance and AI ethics are identical concepts with different names

60 / 60