QCM questions (AI Regulatory Avalanche of Flashcards)

Question 1

Q

Which of the following is a principle of Privacy by Design?
A) Privacy should be an afterthought in AI system development
B) AI systems should not require human intervention
C) Privacy should be embedded into the system by default
D) Privacy concerns should be addressed only when users complain

Answer

A

Correct Answer: C) Privacy should be embedded into the system by default

Explanation: Privacy by Design ensures that privacy safeguards are proactively built into AI systems as a core feature, rather than being added later.

Question 2

Q

What defines a High-Risk AI System under the EU AI Act?
A) AI that has a significant impact on safety or fundamental rights
B) AI models trained using deep reinforcement learning
C) Open-source AI models available for general use
D) AI used in experimental academic research

Answer

A

A) AI that has a significant impact on safety or fundamental rights
Explanation: High-risk AI systems must comply with strict transparency and risk management requirements.

Question 3

Q

How does the EU AI Act ensure transparency in AI decision-making?
A. By mandating that all AI-driven decisions be reviewed by humans
B. By requiring AI deployers to provide clear documentation on how AI systems reach decisions
C. By banning all AI models that are not explainable
D. By allowing AI systems to operate freely without disclosure requirements

Answer

A

Correct Answer: B. By requiring AI deployers to provide clear documentation on how AI systems reach decisions

Explanation: Transparency measures ensure that AI-driven decisions can be explained and challenged, especially in high-risk applications.

Question 4

Q

Which metric is commonly used to measure the explainability of an AI model?
A) Accuracy
B) SHAP (Shapley Additive Explanations)
C) Precision
D) F1-score

Answer

A

B) SHAP (Shapley Additive Explanations)
Explanation: SHAP values help quantify feature importance and explain AI model decisions.

Question 5

Q

According to EO 14110, AI developers must:
A) Share safety test results with the U.S. government
B) Obtain a federal AI safety license before deployment
C) Implement carbon neutrality standards for AI infrastructure
D) Disclose training datasets used in AI systems

Answer

A

A) Share safety test results with the U.S. government
Explanation: The order emphasizes AI risk management, requiring transparency in AI safety testing.

Question 6

Q

Which of the following is not covered under copyright law?
A) Software source code
B) Facts and ideas
C) Literary works
D) Architectural designs

Answer

A

B) Facts and ideas
Explanation: Copyright law protects the expression of ideas, not the ideas themselves. Facts and systems are not copyrightable.

Question 7

Q

What is the main function of adversarial testing in AI security?
A) To train AI models to resist attacks and manipulation
B) To improve AI-generated images
C) To increase AI system complexity
D) To validate AI model accuracy

Answer

A

Correct Answer: A) To train AI models to resist attacks and manipulation

Explanation: Adversarial testing evaluates AI robustness by simulating attacks, helping detect vulnerabilities before real-world deployment.

Question 8

Q

Which data format is easiest to analyze using traditional SQL-based tools?
A) Structured data
B) Unstructured data
C) Semi-structured data
D) Streaming data

Answer

A

A) Structured data
Explanation: Structured data, stored in relational databases with tables and columns, is ideal for SQL-based analysis.

Question 9

Q

Which of the following is an essential aspect of AI governance?
A) Maximizing model accuracy
B) Ensuring AI compliance with legal and ethical guidelines
C) Selecting the most efficient deep learning model
D) Reducing data storage costs

Answer

A

B) Ensuring AI compliance with legal and ethical guidelines
Explanation: AI governance focuses on risk assessment, accountability, and regulatory compliance.

Question 10

Q

Which AI paradigm best describes reinforcement learning?
A) AI learns from labeled data
B) AI extracts patterns from data without supervision
C) AI learns through rewards and penalties
D) AI generates data rather than classifies it

Answer

A

Correct Answer: C) AI learns through rewards and penalties

Explanation: Reinforcement learning is based on feedback loops, where AI takes actions and receives rewards or punishments, allowing it to optimize decision-making over time.

Question 11

Q

How does AI explainability relate to compliance with both proprietary AI governance and the NIST AI RMF?
A) The NIST AI RMF mandates explainability, while proprietary AI models are often black-box systems
B) Explainability is only relevant for AI models used in the financial sector
C) The EU AI Act does not require AI explainability
D) Proprietary AI models are always fully transparent

Answer

A

A) The NIST AI RMF mandates explainability, while proprietary AI models are often black-box systems
Explanation: The NIST AI RMF promotes explainability as a best practice, whereas proprietary AI models often lack transparency, complicating risk assessment and compliance.

Question 12

Q

What is a key governance challenge in auditing proprietary AI models?
A) The algorithms are too simple to require auditing
B) The model owners do not allow external audits
C) Proprietary models are always safe to deploy
D) Open-source models face stricter regulations

Answer

A

B) The model owners do not allow external audits
Explanation: Proprietary AI providers often limit external audits, making it difficult to evaluate bias, safety, and ethical concerns.

Question 13

Q

How does the EU AI Act’s risk-based approach impact AI compliance for proprietary AI models?
A) Proprietary AI models must be assessed for risk level and comply with relevant transparency and accountability obligations
B) The EU AI Act does not classify AI risks
C) Proprietary AI models are always classified as low-risk
D) The EU AI Act does not regulate proprietary AI models

Answer

A

A) Proprietary AI models must be assessed for risk level and comply with relevant transparency and accountability obligations
Explanation: The EU AI Act enforces obligations based on an AI system’s risk level, including proprietary models used in high-risk applications.

Question 14

Q

Which industry is most likely to be affected by AI-driven job automation?
A. Healthcare
B. Finance
C. Manufacturing
D. Retail

Answer

A

Correct Answer: C. Manufacturing

Explanation: AI-driven automation significantly impacts manufacturing jobs by replacing repetitive, manual tasks with intelligent robotic systems.

Question 15

Q

Which of the following is not a typical AI deployment challenge?
A) Latency issues
B) Model packaging compatibility
C) Unlimited scalability
D) Security vulnerabilities

Answer

A

C) Unlimited scalability
Explanation: AI deployment often faces scalability limits, latency issues, and security risks—but unlimited scalability is not a typical challenge.

Question 16

Q

Which proprietary AI model sparked debates over intellectual property and fair use?
A) OpenAI’s GPT
B) Google’s BERT
C) Meta’s LLaMA
D) IBM Watson

Answer

A

A) OpenAI’s GPT
Explanation: GPT models have faced lawsuits regarding their use of copyrighted data in training without explicit consent.

Question 17

Q

Which of the following best defines a foundation model?
A) A neural network trained on a massive dataset for broad applicability
B) A traditional rule-based AI system
C) An AI system specifically designed for real-time decision-making
D) An AI model optimized for small-scale data analysis

Answer

A

Correct Answer: A) A neural network trained on a massive dataset for broad applicability

Explanation: Foundation models, such as large language models, are trained on extensive datasets and can be fine-tuned for specific applications.

Question 18

Q

Which type of AI model architecture is best suited for image recognition tasks?
A) Recurrent Neural Networks (RNN)
B) Convolutional Neural Networks (CNN)
C) Decision Trees
D) Reinforcement Learning

Answer

A

B) Convolutional Neural Networks (CNN)
Explanation: CNNs are specialized for processing image data by detecting patterns through convolutional and pooling layers.

Question 19

Q

Which ‘V’ of big data refers to the accuracy and reliability of data?
A) Variety
B) Velocity
C) Veracity
D) Volume

Answer

A

C) Veracity
Explanation: The 5 Vs of big data include Variety (data types), Velocity (processing speed), Volume (amount), Veracity (accuracy), and Value (usefulness).

Question 20

Q

What is the role of the Market Surveillance Authority (MSA)?
A) Monitor AI compliance and enforce penalties at the national level
B) Develop AI governance guidelines for private companies
C) Approve all AI model releases before deployment
D) Conduct AI research for regulatory bodies

Answer

A

A) Monitor AI compliance and enforce penalties at the national level
Explanation: MSAs enforce AI Act compliance within their respective EU member states.

Question 21

Q

Which of the following best defines the OECD AI Transparency Principle?
A) AI systems should explain their decision-making processes
B) AI models should be open-source
C) AI should be regulated by a single international body
D) AI-generated content should always be labeled

Answer

A

A) AI systems should explain their decision-making processes
Explanation: Transparency ensures that users and stakeholders can understand and trust AI outputs.

Question 22

Q

How does the principle of AI accountability in the NIST AI RMF align with compliance requirements under the EU AI Act?
A) Both frameworks mandate that AI models must be decentralized
B) AI accountability requires clear roles, governance structures, and compliance measures in both frameworks
C) The EU AI Act does not include accountability as a regulatory principle
D) AI accountability is only necessary for AI models used in cybersecurity

Answer

A

B) AI accountability requires clear roles, governance structures, and compliance measures in both frameworks
Explanation: Both frameworks emphasize governance structures that ensure AI accountability, requiring clear policies and compliance measures.

Question 23

Q

Under GDPR, which of the following is considered a legal basis for data processing?
A) User consent
B) Data ownership transfer
C) AI system autonomy
D) Automated bias detection

Answer

A

Correct Answer: A) User consent

Explanation: Organizations must obtain valid user consent or rely on another legal basis (e.g., contract necessity or legal obligation) before processing personal data.

Question 24

Q

Under the EU AI Act, what penalty can be imposed for deploying a prohibited AI system?
A. No penalty, as AI development is encouraged
B. Up to €35 million or 7% of global annual turnover
C. A written warning with a 5-year grace period
D. Immediate ban without any financial consequences

Answer

A

Correct Answer: B. Up to €35 million or 7% of global annual turnover

Explanation: The EU AI Act imposes significant financial penalties on companies violating AI regulations, particularly for deploying prohibited AI applications.

Question 25

Q

Which authority is responsible for supervising General Purpose AI (GPAI) models at the EU level?
A) European AI Office
B) Market Surveillance Authority
C) European Patent Office
D) OECD AI Governance Committee

Answer

A

A) European AI Office
Explanation: The European AI Office oversees compliance and risk assessments for GPAI models.

Question 26

Q

What is a shared governance challenge for proprietary AI models and high-risk AI applications under the EU AI Act?
A) Both require clear accountability structures, transparency, and bias mitigation
B) Proprietary AI models are automatically classified as low-risk
C) The EU AI Act does not regulate high-risk AI models
D) Proprietary AI models are exempt from all compliance requirements

Answer

A

A) Both require clear accountability structures, transparency, and bias mitigation
Explanation: Both proprietary AI and high-risk AI applications must meet transparency and accountability standards to align with governance frameworks.

Question 27

Q

NYC Local Law 144 requires that AI-based hiring tools:
A) Undergo bias audits before deployment
B) Be entirely transparent about their decision-making process
C) Receive federal certification before use
D) Be manually reviewed by human HR professionals before making hiring decisions

Answer

A

A) Undergo bias audits before deployment
Explanation: Local Law 144 mandates that AI hiring systems be audited for bias and that employers notify candidates when such tools are used in hiring decisions.

Question 28

Q

Which governance measure helps limit liability when deploying AI models?
A) Open-sourcing all AI models
B) Implementing technical documentation, AUPs, and compliance reviews
C) Using only on-premise AI infrastructure
D) Allowing AI models to function independently without audits

Answer

A

B) Implementing technical documentation, AUPs, and compliance reviews
Explanation: Organizations limit liability by maintaining detailed documentation, Acceptable Use Policies (AUPs), and compliance checks.

Question 29

Q

What is a key requirement for deployers using High-Risk AI Systems?
A) Provide real-time access to system logs to all users
B) Conduct a Fundamental Rights Impact Assessment (FRIA)
C) Open-source all AI models
D) Ensure AI models do not use deep learning

Answer

A

B) Conduct a Fundamental Rights Impact Assessment (FRIA)
Explanation: Deployers of high-risk AI systems must conduct FRIA assessments to evaluate risks to fundamental rights.

Question 30

Q

What is the primary risk associated with black-box AI models?
A) They are too slow for real-world deployment
B) Their decision-making processes are not interpretable
C) They cannot process unstructured data
D) They require constant internet connectivity

Answer

A

Correct Answer: B) Their decision-making processes are not interpretable

Explanation: Black-box AI models, such as deep neural networks, lack transparency, making it difficult to understand how decisions are made.

Question 31

Q

Which of the following is a physical AI risk control?
A. Secure data centers for AI model storage
B. Algorithmic fairness audits
C. Ethical AI policies
D. Bias impact assessments

Answer

A

Correct Answer: A. Secure data centers for AI model storage

Explanation: Physical controls include secure locations, locked data centers, and restricted access areas to prevent unauthorized AI tampering.

Question 32

Q

How do the EU AI Act and NIST AI RMF approach AI accountability differently?
A) The EU AI Act mandates accountability measures, while NIST AI RMF provides voluntary guidelines
B) Both frameworks enforce strict penalties for non-compliance
C) NIST AI RMF requires government audits, while the EU AI Act does not
D) AI accountability is only relevant to proprietary AI models

Answer

A

A) The EU AI Act mandates accountability measures, while NIST AI RMF provides voluntary guidelines
Explanation: The EU AI Act enforces legally binding accountability rules, while NIST AI RMF provides flexible, voluntary best practices for AI risk management.

Question 33

Q

How do the risk classification systems of the EU AI Act and NIST AI RMF fundamentally differ?
A) The EU AI Act enforces legally binding risk categories, while NIST AI RMF provides adaptable risk frameworks
B) Both frameworks classify AI risks in the exact same way
C) NIST AI RMF does not include AI risk classifications
D) The EU AI Act does not regulate AI risk

Answer

A

A) The EU AI Act enforces legally binding risk categories, while NIST AI RMF provides adaptable risk frameworks
Explanation: The EU AI Act categorizes AI into mandatory risk levels, while NIST AI RMF allows for flexible, organization-specific risk management strategies.

Question 34

Q

What key obligation does the EU AI Act impose on deployers of high-risk AI?
A. They must delete AI models every six months
B. They must implement and document human oversight of AI systems
C. They must prevent all users from interacting with AI decision-making tools
D. They must make AI systems open-source

Answer

A

Correct Answer: B. They must implement and document human oversight of AI systems

Explanation: The EU AI Act mandates human intervention where necessary to ensure fairness and accountability in high-risk AI decision-making.

Question 35

Q

Which risk is associated with AI data drift?
A) AI model becomes more accurate over time
B) AI outputs become unpredictable due to changing input distributions
C) AI models self-correct errors in real-time
D) AI models only work in real-time applications

Answer

A

B) AI outputs become unpredictable due to changing input distributions
Explanation: Data drift occurs when the statistical properties of input data change, leading to unreliable AI performance.

Question 36

Q

What is the primary characteristic that differentiates Artificial Narrow Intelligence (ANI) from Artificial General Intelligence (AGI)?
A) ANI can generalize knowledge across multiple domains, AGI cannot
B) ANI performs a single function within defined constraints, while AGI exhibits human-level intelligence
C) ANI is entirely theoretical, while AGI is already widely deployed
D) ANI and AGI have the same capabilities, but AGI is more efficient

Answer

A

Correct Answer: B) ANI performs a single function within defined constraints, while AGI exhibits human-level intelligence

Explanation: ANI is designed for specific tasks, such as facial recognition or language translation, whereas AGI is capable of reasoning, problem-solving, and adapting across various domains like a human.

Question 37

Q

Which AI deployment approach is best suited for real-time AI inference?
A) Batch processing
B) Edge computing
C) Data lake storage
D) Offline processing

Answer

A

B) Edge computing
Explanation: Edge computing enables real-time inference by running AI models directly on local devices.

Question 38

Q

How does AI governance under the NIST AI RMF align with compliance requirements in the EU AI Act?
A) Both frameworks emphasize risk management and accountability
B) The EU AI Act requires open-source AI models, while NIST AI RMF does not
C) NIST AI RMF does not address AI governance principles
D) The EU AI Act does not require AI accountability

Answer

A

A) Both frameworks emphasize risk management and accountability
Explanation: The NIST AI RMF and EU AI Act share governance principles, requiring organizations to define AI risks, implement monitoring mechanisms, and ensure ethical accountability.

Question 39

Q

What is a major security concern with proprietary AI models?
A) They consume too much computational power
B) They can contain undisclosed vulnerabilities that lead to adversarial attacks
C) They can never be updated after deployment
D) They cannot be used in regulated industries

Answer

A

B) They can contain undisclosed vulnerabilities that lead to adversarial attacks
Explanation: Since proprietary AI models are closed-source, security vulnerabilities may exist without public awareness or scrutiny.

Question 40

Q

What is the main risk of using synthetic data in AI model training?
A) It is legally restricted in most jurisdictions
B) It lacks diversity compared to real-world data
C) It always produces biased AI models
D) It requires deep learning models to be effective

Answer

A

B) It lacks diversity compared to real-world data
Explanation: Synthetic data may not fully represent the diversity of real-world data, leading to biased or inaccurate AI models.

Question 41

Q

Which AI risk assessment framework focuses on continuous evaluation and adaptation?
A) ISO 9001
B) NIST AI RMF
C) IEEE 7000-21
D) SOC 2 Compliance

Answer

A

B) NIST AI RMF
Explanation: The NIST AI Risk Management Framework (AI RMF) promotes continuous risk evaluation and mitigation in AI systems.

Question 42

Q

How do proprietary AI models pose challenges in compliance with the NIST AI RMF and the EU AI Act?
A) Proprietary AI models always comply with both frameworks
B) Lack of transparency in proprietary models makes it difficult to assess risks and biases
C) Proprietary models cannot be used in regulated industries
D) Proprietary AI models must be open-source to comply with both frameworks

Answer

A

B) Lack of transparency in proprietary models makes it difficult to assess risks and biases
Explanation: Proprietary models often do not disclose their training data or decision-making processes, creating challenges for compliance with transparency and risk assessment requirements in both frameworks.

Question 43

Q

The case “Silverman v. OpenAI” primarily revolved around which legal question?
A) Whether AI-generated content is protected under copyright law
B) Whether training AI models using copyrighted books without permission constitutes infringement
C) Whether AI-generated books can receive fair use protection
D) Whether AI systems can autonomously apply fair use defenses

Answer

A

B) Whether training AI models using copyrighted books without permission constitutes infringement
Explanation: The plaintiffs claimed OpenAI used copyrighted books to train its AI models without consent, but the court ruled that training on copyrighted data did not automatically constitute infringement.

Question 44

Q

How does the EU AI Act’s risk-based approach impact the deployment of General Purpose AI (GPAI)?
A) GPAI models require additional compliance measures because of their wide applicability
B) The EU AI Act classifies GPAI models as low-risk by default
C) GPAI models are not subject to any AI regulation
D) The EU AI Act does not address GPAI compliance

Answer

A

A) GPAI models require additional compliance measures because of their wide applicability
Explanation: The EU AI Act enforces transparency and risk management requirements on GPAI due to their broad integration across multiple applications.

Question 45

Q

Which of the following is NOT a primary reason to pursue AIGP certification?
A. To gain expertise in AI governance
B. To understand AI model development in-depth
C. To advance career opportunities
D. To support responsible AI deployment

Answer

A

Correct Answer: B. To understand AI model development in-depth

Explanation: AIGP certification focuses on AI governance, compliance, and risk management rather than technical AI development itself.

Question 46

Q

Which of the following is a key principle in the NIST AI RMF?
A) AI systems should be entirely automated
B) AI risk management should be flexible and context-dependent
C) AI regulation should be uniform across all industries
D) AI models should not require any human oversight

Answer

A

B) AI risk management should be flexible and context-dependent
Explanation: The NIST AI RMF emphasizes that risk management approaches should adapt based on the AI system’s context and use case.

Question 47

Q

Which case study best exemplifies reinforcement learning AI under OECD’s framework?
A) AlphaGo
B) ChatGPT-4
C) IBM Watson
D) Google Translate

Answer

A

A) AlphaGo
Explanation: AlphaGo used reinforcement learning to optimize gameplay strategies, aligning with the framework’s AI model classification.

Question 48

Q

What is a common governance challenge for both proprietary AI models and high-risk AI applications under the EU AI Act?
A) The requirement for AI models to be open-source
B) The need for robust documentation, bias audits, and transparency reporting
C) The exemption of proprietary AI from AI regulations
D) The requirement to train AI models only on European datasets

Answer

A

B) The need for robust documentation, bias audits, and transparency reporting
Explanation: Both proprietary AI models and high-risk AI applications must adhere to compliance measures that involve transparency, accountability, and bias mitigation.

Question 49

Q

Which NIST framework aims to ensure AI systems are safe, explainable, and fair?
A) NIST AI RMF
B) NIST Cybersecurity Framework
C) IEEE 7000-21
D) ISO 22989:2022

Answer

A

A) NIST AI RMF
Explanation: The NIST AI Risk Management Framework (AI RMF) provides a structure to ensure AI safety, fairness, and transparency.

Question 50

Q

How does the concept of AI bias mitigation in the NIST AI RMF relate to high-risk AI applications under the EU AI Act?
A) The EU AI Act requires bias mitigation only for proprietary AI models
B) The NIST AI RMF provides guidelines for detecting and mitigating bias, while the EU AI Act mandates bias audits for high-risk AI systems
C) AI bias is not addressed by either framework
D) The EU AI Act requires all AI models to be bias-free before deployment

Answer

A

B) The NIST AI RMF provides guidelines for detecting and mitigating bias, while the EU AI Act mandates bias audits for high-risk AI systems
Explanation: The NIST AI RMF suggests best practices for bias detection, while the EU AI Act legally requires audits and documentation for high-risk AI models.

Question 51

Q

Which principle of Fair Information Practices aligns with the concept of ‘data minimization’?
A) Use Limitation
B) Data Quality and Relevance
C) Collection Limitation
D) Safeguards

Answer

A

Correct Answer: C) Collection Limitation

Explanation: Collection Limitation (also known as Data Minimization) ensures that only necessary data is collected and retained for as long as needed.

Question 52

Q

Which U.S. executive order mandates AI developers to report safety test results to the government?
A) EO 14091
B) EO 14110
C) EO 13960
D) EO 13769

Answer

A

B) EO 14110
Explanation: Executive Order 14110 strengthens AI safety by requiring developers to share risk assessments with the federal government.

Question 53

Q

How does the EU AI Act define ‘General-Purpose AI Models’ (GPAI)?
A. AI systems that are designed for a single, highly specialized task
B. AI models trained on large datasets that can perform a broad range of tasks
C. AI applications that are used exclusively for governmental purposes
D. AI models that are exempt from all transparency requirements

Answer

A

Correct Answer: B. AI models trained on large datasets that can perform a broad range of tasks

Explanation: GPAI models, also known as foundation models, are trained on extensive datasets and can be adapted for multiple downstream applications.

Question 54

Q

What is the maximum fine for GDPR violations?
A) €10 million or 2% of global turnover
B) €20 million or 4% of global turnover
C) €30 million or 6% of global turnover
D) No financial penalty, only a warning

Answer

A

Correct Answer: B) €20 million or 4% of global turnover

Explanation: GDPR imposes strict penalties for non-compliance, with maximum fines reaching 4% of a company’s global annual revenue or €20 million, whichever is higher.

Question 55

Q

Which of the following is not a key component of AI project planning?
A) Business problem
B) Mission
C) System architecture
D) Data considerations

Answer

A

C) System architecture
Explanation: AI project planning includes defining the business problem, mission, gaps, data considerations, and governance. System architecture is determined in the design phase.

Question 56

Q

Which of the following AI liability models applies no-fault responsibility to developers?
A) Fault-based liability
B) Negligence-based liability
C) Strict liability
D) Common carrier liability

Answer

A

C) Strict liability
Explanation: Under strict liability, AI developers may be held responsible for harm caused by their products regardless of intent or negligence.

Question 57

Q

Under GDPR, when is a DPIA required?
A) When deploying AI models with minimal risk
B) When processing high-risk personal data
C) Only when requested by regulators
D) Only for organizations operating outside the EU

Answer

A

Correct Answer: B) When processing high-risk personal data

Explanation: DPIAs are mandatory under GDPR for high-risk AI systems, ensuring potential privacy harms are evaluated before data processing begins.

Question 58

Q

What is the primary function of the EU AI Act’s Articles?
A) Provide supplementary technical specifications
B) Explain the reasoning behind legislative provisions
C) Establish substantive rules, rights, and obligations
D) Define AI governance for member states

Answer

A

C) Establish substantive rules, rights, and obligations
Explanation: Articles in the EU AI Act set out legally binding rules and obligations that directly affect AI operators.

Question 59

Q

Under the EU AI Act, what is required of AI providers placing high-risk AI systems on the EU market?
A. No specific requirements, as AI is self-regulated
B. Conformity assessments, risk management, and post-market monitoring
C. Full government oversight and pre-approval of all AI systems
D. Exemptions from all liability if AI models fail

Answer

A

Correct Answer: B. Conformity assessments, risk management, and post-market monitoring

Explanation: High-risk AI providers must conduct assessments to ensure compliance, track risks, and continuously monitor the AI’s impact on users and society.

Question 60

Q

How does the principle of explainable AI (XAI) in the NIST AI RMF impact compliance with proprietary AI governance?
A) Explainability is not relevant for proprietary AI models
B) Proprietary AI models must implement XAI techniques to meet transparency requirements
C) NIST AI RMF prohibits the use of black-box AI models
D) The EU AI Act does not require explainability for any AI models

Answer

A

B) Proprietary AI models must implement XAI techniques to meet transparency requirements
Explanation: The NIST AI RMF emphasizes the importance of explainability, which helps ensure compliance with transparency and accountability regulations.

Question 61

Q

Which U.S. Executive Order establishes guidelines for AI safety and security?
A) Executive Order 14110
B) Executive Order 13960
C) Executive Order 13769
D) Executive Order 14091

Answer

A

A) Executive Order 14110
Explanation: EO 14110 mandates AI safety measures, requiring companies to share test results and align AI development with national security priorities.

Question 62

Q

NYC Local Law 144 requires that AI-based hiring tools:
A) Undergo bias audits before deployment
B) Be entirely transparent about their decision-making process
C) Receive federal certification before use
D) Be manually reviewed by human HR professionals before making hiring decisions

Answer

A

A) Undergo bias audits before deployment
Explanation: Local Law 144 mandates that AI hiring systems be audited for bias and that employers notify candidates when such tools are used in hiring decisions.

Question 63

Q

Under the Digital Services Act (DSA), what requirement must AI-powered recommender systems comply with?
A) They must avoid collecting personal data for targeted advertising.
B) They must clearly inform users how recommendations are generated.
C) They must provide an alternative manual content ranking system.
D) They must be certified by the European Data Protection Board.

Answer

A

B) They must clearly inform users how recommendations are generated.
Explanation: The DSA enforces transparency requirements for recommender systems, ensuring users understand how content is prioritized and displayed.

Question 64

Q

What does ‘Fairness in AI’ primarily aim to address?
A) Ensuring AI models operate without bias or discrimination
B) Reducing AI system efficiency
C) Allowing unrestricted AI decision-making
D) Preventing AI governance regulations

Answer

A

Correct Answer: A) Ensuring AI models operate without bias or discrimination

Explanation: Fairness in AI ensures that models do not reinforce biases, protecting users from discriminatory outcomes in hiring, finance, healthcare, and law enforcement.

Answer 65

A

B) Ensure compliance on behalf of a non-EU provider
Explanation: Authorized Representatives act as the compliance liaison for AI providers outside the EU.

Answer 66

A

B) Packages code, dependencies, and configurations for easy deployment
Explanation: Containers ensure AI models are portable and compatible across multiple environments.

Answer 67

A

B) Evaluating AI accuracy and detecting irregular decisions
Explanation: Continuous monitoring involves tracking model performance, detecting deviations, and mitigating risks such as data drift.

Answer 68

A

B) The lack of external auditing for proprietary AI models makes continuous risk monitoring more difficult
Explanation: Continuous risk monitoring is essential, but proprietary AI models often restrict access, making ongoing evaluations harder.

Answer 69

A

Correct Answer: A) Generative models create new data, while discriminative models classify existing data

Explanation: Generative models, such as GPT and GANs, generate new samples from learned distributions, while discriminative models classify or separate data into categories.

Answer 70

A

B) Thaler v. Vidal (2023)
Explanation: The “Thaler v. Vidal” case confirmed that only humans can be named inventors on a patent, reinforcing the idea that AI cannot hold copyright or patent rights under U.S. law.

Answer 71

A

B) GPAI models require additional compliance measures because they can be integrated into multiple applications
Explanation: The EU AI Act introduces specific regulations for GPAI due to their broad applicability, while the NIST AI RMF provides adaptable risk management strategies.

Answer 72

A

Correct Answer: B) By adding statistical noise to data to protect individual privacy

Explanation: Differential privacy prevents AI models from learning identifiable patterns by adding controlled randomness, improving privacy protection.

Answer 73

A

Correct Answer: B. They must ensure human oversight and compliance with transparency rules

Explanation: Deployers must monitor AI decisions, ensure human oversight where required, and comply with transparency obligations to protect users’ rights.

Answer 74

A

B) A legal document outlining how an AI model can be used
Explanation: Proprietary AI models require license agreements that define usage rights, limitations, and compliance requirements.

Answer 75

A

A) AI models evolve unpredictably over time
Explanation: One key challenge in AI risk management is tracking and adapting to emergent risks as AI systems evolve.

Answer 76

A

B) Integration complexity and API management
Explanation: Embedding AI requires managing APIs, data flows, and ensuring compatibility with application environments.

Answer 77

A

C) Consumer protection in the financial sector, including AI-driven credit decisions
Explanation: The CFPB ensures transparency and fairness in financial services, including AI-driven lending models and credit scoring.

Answer 78

A

Correct Answer: B) Ann Cavoukian

Explanation: Ann Cavoukian, former Information and Privacy Commissioner of Ontario, developed Privacy by Design to proactively integrate privacy into system design and governance policies.

Answer 79

A

B) Transparency, documentation, and human oversight
Explanation: Both proprietary AI models and high-risk AI applications must implement transparency, documentation, and accountability structures to comply with governance frameworks.

Answer 80

A

B) The lack of access to model architecture and training data makes it harder to verify fairness and risk
Explanation: Proprietary AI models often restrict access to their internal workings, making it difficult to ensure compliance with explainability and risk management standards.

Answer 81

A

Correct Answer: B) Artificial General Intelligence (AGI)

Explanation: AGI exhibits reasoning, learning, and adaptability across diverse fields, similar to human intelligence.

Answer 82

A

C) A specific explanation of the factors leading to the adverse decision
Explanation: The ECOA requires creditors to explain specific reasons for denying credit applications, even when using AI, to prevent discrimination and ensure transparency.

Answer 83

A

A) The lack of transparency in proprietary AI models makes it difficult to assess their risk level
Explanation: Proprietary AI models do not always disclose their internal mechanisms, making it harder to classify them under the EU AI Act’s predefined risk levels.

Answer 84

A

A) It encourages risk assessment methodologies, which are often difficult to implement due to proprietary AI’s closed nature
Explanation: The ‘Measure’ function in NIST AI RMF focuses on risk assessment, but proprietary AI models often limit access to necessary data, complicating compliance.

Answer 85

A

Correct Answer: A) Transparency

Explanation: Transparency ensures that AI models explain their decision-making processes, allowing stakeholders to evaluate fairness, accountability, and compliance.

Answer 86

A

A) Measure
Explanation: The ‘Measure’ function focuses on developing risk assessment methods and evaluation strategies for AI models.

Answer 87

A

C) It requires significant pre-processing and transformation
Explanation: Unstructured data, such as images and text, must be formatted and labeled before AI models can analyze it.

Answer 88

A

C) AI for customer service chatbots
Explanation: Customer service AI is considered low-risk, while social scoring and manipulative AI are banned.

Answer 89

A

A) Their adaptability allows them to be used in multiple applications, making risk assessments more complex
Explanation: Because GPAI can be applied in various domains, ensuring compliance with transparency and accountability standards becomes challenging.

Answer 90

A

B) AI providers must conduct risk assessments, maintain documentation, and ensure model transparency
Explanation: Both proprietary AI governance and high-risk AI applications require compliance with transparency, documentation, and risk management obligations.

Answer 91

A

A) Differential Privacy
Explanation: Differential privacy adds noise to data to protect individual privacy while maintaining useful insights.

Answer 92

A

Correct Answer: B. Real-time biometric surveillance without explicit legal justification

Explanation: Real-time biometric identification in public spaces is banned unless strictly necessary for national security or law enforcement in limited cases.

Answer 93

A

Correct Answer: B. To verify that AI systems comply with EU regulations before being placed on the market

Explanation: Importers must ensure AI models meet EU compliance requirements before distribution and are held accountable for regulatory breaches.

Answer 94

A

D) Explain
Explanation: The four core functions of the NIST AI RMF are Map, Measure, Manage, and Govern. ‘Explain’ is not a separate function but falls under transparency efforts.

Answer 95

A

Correct Answer: A) Organizations take responsibility for ensuring compliance with data protection policies

Explanation: Accountability ensures organizations implement data governance policies, adhere to regulations, and enforce privacy protections.

Answer 96

A

B) Ensure AI is safe, trustworthy, and respects fundamental rights
Explanation: The EU AI Act seeks to balance AI innovation with fundamental rights protections.

Answer 97

A

B) AI-generated works cannot be copyrighted because they lack human authorship.
Explanation: U.S. copyright law requires human authorship. Cases such as “Thaler v. Vidal (2023)” confirm that AI cannot be named as an inventor, and the U.S. Copyright Office has repeatedly denied AI-generated works copyright protection.

Answer 98

A

A) Data drift
Explanation: Continuous monitoring ensures that AI models maintain accuracy by detecting data drift and adapting to changes over time.

Answer 99

A

D) Algorithmic Bias Compensation
Explanation: While OECD AI principles emphasize fairness and non-discrimination, they do not explicitly require AI to actively compensate for bias.

Answer 100

A

A) Data poisoning attacks
Explanation: Data poisoning attacks occur when malicious data is introduced to manipulate AI model behavior.

Answer 101

A

B) AI operates autonomously, making it difficult to attribute harm to a specific party.
Explanation: Because AI systems function independently, determining responsibility—whether it’s the developer, the user, or the AI itself—is legally complex.

Answer 102

A

Correct Answer: C) AI-generated policies

Explanation: AI governance frameworks are created by human regulators and organizations to ensure AI fairness, accountability, and transparency.

Answer 103

A

B) They provide better security and exclusive access to optimized performance
Explanation: Proprietary models often leverage proprietary data and optimizations, leading to better security and performance advantages for their owners.

Answer 104

A

B) Lack of transparency in decision-making
Explanation: Black-box models, like deep neural networks, make decisions that are difficult to interpret, creating transparency and accountability challenges.

Answer 105

A

Correct Answer: D. AI-generated self-regulatory controls

Explanation: AI governance relies on human-defined risk controls, such as administrative policies, technical measures, and physical safeguards.

Answer 106

A

A) Conduct periodic bias audits and risk assessments
Explanation: AI providers must regularly assess bias, transparency, and risks before deployment.

Answer 107

A

Correct Answer: B. Strict compliance with transparency, fairness, and bias mitigation requirements

Explanation: AI used in law enforcement is subject to high-risk regulations, requiring human oversight, fairness testing, and impact assessments.

Answer 108

A

A) Proprietary AI models often limit external oversight, making risk audits and compliance assessments difficult
Explanation: Many proprietary AI models operate as black boxes, making compliance with AI governance principles more challenging.

Answer 109

A

C) Model explainability
Explanation: Explainability is important but is primarily an AI governance concern, while deployment considerations focus on infrastructure, integration, and access.

Answer 110

A

B) Thaler v. Vidal (2023)
Explanation: The “Thaler v. Vidal” case confirmed that only humans can be named inventors on a patent, reinforcing the idea that AI cannot hold copyright or patent rights under U.S. law.

Answer 111

A

A) General Purpose AI (GPAI)
Explanation: GPAI, also called Foundation Models, are trained on large-scale data and exhibit broad generalization capabilities.

Answer 112

A

D) Transparency and Explainability
Explanation: The OECD AI Principles include this category to ensure AI systems are understandable and their decision-making processes are clear.

Answer 113

A

Correct Answer: A) Data Quality and Accuracy

Explanation: Data Quality and Accuracy require organizations to maintain updated and correct personal data to prevent errors and misinformation.

Answer 114

A

D) Model accuracy
Explanation: AI project scope includes Impact, Effort, and Fit, whereas accuracy is an evaluation metric in the development phase.

Answer 115

A

B) Council of Europe AI Convention
Explanation: This treaty aims to govern AI development while ensuring fundamental rights, democracy, and rule of law.

Answer 116

A

Correct Answer: B. GDPR

Explanation: The General Data Protection Regulation (GDPR) mandates strict AI data privacy protections and user rights regarding AI-driven decisions.

Answer 117

A

B) Identifying whether an email is spam or not
Explanation: Classification models categorize inputs into discrete labels, such as ‘spam’ or ‘not spam’.

Answer 118

A

D) Mandatory open-source AI models
Explanation: ISO/IEC 42001:2023 focuses on risk management and governance, but it does not require AI models to be open-source.

Answer 119

A

C) Provide transparency about content moderation and recommendation algorithms
Explanation: The DSA aims to enhance accountability of online platforms by ensuring transparency in content recommendation and moderation policies.

Answer 120

A

B) The EU AI Act mandates continuous monitoring for high-risk AI systems, aligning with the NIST AI RMF’s emphasis on ongoing risk assessment
Explanation: Both frameworks recognize the importance of continuously assessing AI risks and performance post-deployment to ensure compliance and safety.

Answer 121

A

C) Limited Risk
Explanation: AI chatbots fall under the Limited Risk category, requiring only transparency obligations.

Answer 122

A

B) Implementing a risk assessment framework
Explanation: AI governance involvesestablishing policies, overseeing risk management, and ensuring regulatory compliance.

Answer 123

A

A) Proprietary AI models often do not disclose risk factors, making classification and compliance difficult
Explanation: Proprietary AI models lack transparency, complicating their classification under the EU AI Act’s risk-based approach.

Answer 124

A

B) The EU AI Act focuses on regulatory compliance, while NIST AI RMF provides voluntary risk management guidance
Explanation: The EU AI Act classifies AI systems into different risk levels with mandatory compliance, while the NIST AI RMF provides flexible risk management principles.

Answer 125

A

B) Changes in real-world data affecting AI accuracy
Explanation: Data drift occurs when input data changes over time, reducing model reliability.

Answer 126

A

A) Risk mitigation should be a continuous, iterative process
Explanation: The NIST AI RMF stresses the importance of ongoing monitoring, assessment, and refinement of AI risk management strategies.

Answer 127

A

Correct Answer: C) AI Autonomy

Explanation: Fair Information Practices (FIPs) focus on privacy, security, and fairness. AI autonomy is not part of FIPs, as these principles emphasize human oversight and accountability.

Answer 128

A

A) Proprietary AI models often restrict access to internal operations, making risk monitoring difficult
Explanation: Proprietary models are typically closed-source, which complicates efforts to assess and monitor AI risks as required by NIST AI RMF.

Answer 129

A

Correct Answer: D. All of the above

Explanation: AIGP certification provides comprehensive knowledge of AI governance, including legal, ethical, and operational aspects.

Answer 130

A

Correct Answer: B) To identify potential risks and mitigation strategies for AI deployment

Explanation: AI risk assessments evaluate factors such as bias, fairness, and security vulnerabilities, ensuring responsible AI deployment.

Answer 131

A

B) Data Controller
Explanation: AI operators include Providers, Deployers, Importers, Distributors, and Authorized Representatives, but ‘Data Controller’ is a GDPR term.

Answer 132

A

Correct Answer: C. The complexity of AI models making decisions difficult to explain

Explanation: Complex AI systems, such as deep learning models, often operate as ‘black boxes,’ making their decision-making processes hard to interpret.

Answer 133

A

C) Open-source licensing
Explanation: While bias audits, security, and explainability are required, AI models do not have to be open-source to comply with regulations.

Answer 134

A

A) The EU AI Act categorizes AI into strict risk levels, while NIST AI RMF provides adaptable risk management strategies
Explanation: The EU AI Act mandates specific obligations based on an AI system’s risk level, while the NIST AI RMF allows organizations to determine how they manage AI risks.

Answer 135

A

B) Conducting algorithmic impact assessments and bias audits
Explanation: Both frameworks emphasize fairness through impact assessments and bias audits to identify and mitigate discriminatory AI behavior.

Answer 136

A

Correct Answer: B. AI-generated chatbots and virtual assistants

Explanation: Limited-risk AI applications require transparency measures, such as informing users that they are interacting with AI but do not face strict regulatory compliance.

Answer 137

A

Correct Answer: B. Using firewalls and encryption to protect AI systems

Explanation: Technical controls ensure AI security through mechanisms such as encryption, firewalls, and adversarial robustness testing.

Answer 138

A

C) Convolutional Neural Networks (CNNs)
Explanation: CNNs are optimized for processing visual data, using convolutional layers to detect spatial hierarchies of features.

Answer 139

A

Correct Answer: C. AI-powered virtual assistants for general consumer use

Explanation: AI virtual assistants fall under limited-risk AI, which only requires transparency measures. High-risk AI applications impact fundamental rights and require stringent compliance measures.

Answer 140

A

Correct Answer: B. IAPP

Explanation: The International Association of Privacy Professionals (IAPP) offers the AIGP certification to professionals specializing in AI governance and compliance.

Answer 141

A

Correct Answer: C. Undergo third-party conformity assessments and ensure compliance with transparency and risk management requirements

Explanation: The EU AI Act mandates that high-risk AI providers submit to external audits and maintain detailed compliance documentation.

Answer 142

A

B) Continuous monitoring and retraining
Explanation: AI drift occurs when data patterns shift over time, requiring constant monitoring and periodic model retraining.

Answer 143

A

B) On-premise deployment
Explanation: On-premise AI ensures full control over data, making it the best option for privacy-sensitive applications.

Answer 144

A

C) Decision trees
Explanation: Classification problems, such as spam detection, are best solved using supervised learning models like decision trees, SVMs, or deep learning classifiers.

Answer 145

A

C) Limited Risk
Explanation: Limited-risk AI systems must comply with transparency requirements but do not face stringent compliance rules like high-risk systems.

Answer 146

A

B) Defining how AI models, dependencies, and configurations are bundled
Explanation: AI deployment packaging ensures compatibility across environments by bundling dependencies and configurations.

Answer 147

A

A) NIST AI RMF emphasizes continuous risk assessment, which is difficult for proprietary AI due to lack of transparency
Explanation: Proprietary AI models often limit access to their internal workings, making compliance with risk management frameworks more challenging.

Answer 148

A

B) Ensuring AI tools comply with consumer protection and fair competition laws
Explanation: The FTC enforces consumer protection and fair competition laws, ensuring AI models do not engage in deceptive practices.

Answer 149

A

Correct Answer: B. The AI deployer or importer within the EU

Explanation: Importers and deployers must ensure that AI systems comply with EU laws before deployment, even if the model was developed outside the EU.

Answer 150

A

A) NIST AI RMF’s ‘Manage’ function encourages organizations to implement continuous risk mitigation strategies, which aligns with the EU AI Act’s high-risk AI obligations
Explanation: Both frameworks emphasize the importance of actively managing and mitigating AI risks through structured processes.

Answer 151

A

Correct Answer: B. Difficulty in monitoring AI compliance across various jurisdictions

Explanation: Since AI systems can be developed outside the EU but deployed within it, enforcing compliance requires coordinated monitoring and oversight mechanisms.

Answer 152

A

B) Provide transparency reports on AI risk assessments
Explanation: The EU AI Act mandates that providers of high-risk AI models disclose risk assessments and mitigation strategies.

Answer 153

A

B) An individual or company that develops or places an AI system on the EU market
Explanation: AI Providers create, modify, or distribute AI models and are subject to the highest regulatory obligations.

Answer 154

A

Correct Answer: B) To ensure AI systems adhere to legal, ethical, and operational standards

Explanation: AI regulatory compliance ensures that AI applications align with legal requirements and ethical principles, protecting users and organizations.

Answer 155

A

A) A risk management framework for AI governance
Explanation: ISO 31000:2018 offers guidelines for risk assessment across industries, including AI.

Answer 156

A

A) It makes auditing, risk assessment, and bias mitigation harder
Explanation: Proprietary models restrict access to decision-making processes, making it difficult to ensure compliance with transparency and fairness requirements.

Answer 157

A

A) API integration
Explanation: REST APIs enable external applications and users to interact with AI models, facilitating broader adoption.

Answer 158

A

B) White House Office of Science and Technology Policy (OSTP)
Explanation: The OSTP introduced the AI Bill of Rights to protect privacy, ensure fairness, and promote AI safety.

Answer 159

A

B) Legally binding, setting out rights and obligations
Explanation: Articles establish substantive legal rules and obligations with direct effect.

Answer 160

A

Correct Answer: C) AI Bill of Rights

Explanation: While the AI Bill of Rights provides ethical guidelines for AI development, it is not a legally binding privacy law like GDPR, CCPA, or BIPA.

Answer 161

A

B) AI system lifecycle stages
Explanation: ISO 42001:2023 provides a structured approach to managing risks across AI development, deployment, and maintenance stages.

Answer 162

A

Correct Answer: B) Data Minimization

Explanation: Data Minimization ensures that organizations collect only the data needed for specific purposes, reducing privacy risks and potential misuse.

Answer 163

A

Correct Answer: C. €35 million or 7% of global turnover

Explanation: The EU AI Act enforces heavy penalties for serious violations, such as deploying prohibited AI systems that threaten fundamental rights.

Answer 164

A

A) Data encryption
Explanation: Encrypting AI models helps prevent unauthorized access and protects sensitive training data.

Answer 165

A

B) The obligation to conduct Fundamental Rights Impact Assessments (FRIA)
Explanation: The ‘Govern’ function of NIST AI RMF focuses on establishing policies and accountability structures, which aligns with the EU AI Act’s requirement for impact assessments in high-risk AI applications.

Answer 166

A

B) Significance, quality, and addition of something new
Explanation: The ‘Pannu Factors’ were established in Pannu v. Iolab (1993) to determine whether a person made a substantial contribution to an invention.

Answer 167

A

Correct Answer: B) Ensuring fairness, accountability, and transparency (FAT)

Explanation: Fairness, accountability, and transparency (FAT) principles guide responsible AI development to prevent discrimination and promote ethical AI use.

Answer 168

A

C) Algorithmic Complexity
Explanation: The framework includes ‘People and Planet,’ ‘Data and Input,’ ‘AI Model,’ ‘Tasks and Output,’ and ‘Economic Context,’ but not ‘Algorithmic Complexity’.

Answer 169

A

B) Any organization or individual using an AI system under their authority
Explanation: Deployers use AI models in operational settings and must ensure responsible usage.

Answer 170

A

B) Strict liability
Explanation: Strict liability applies when AI causes harm, regardless of whether negligence was involved.

Answer 171

A

B) Hallucination
Explanation: Hallucination occurs when AI generates responses that are not based on real or accurate data.

Answer 172

A

A) Cloud
Explanation: Cloud deployment allows for scalability, flexibility, and resource sharing, making it ideal for large-scale AI models.

Answer 173

A

B) Updating training data and fine-tuning the model
Explanation: Retraining AI involves updating datasets and adjusting model parameters to improve accuracy.

Answer 174

A

A) Lack of transparency
Explanation: Proprietary AI models are often ‘black boxes,’ making it difficult to audit or explain their decision-making processes.

Answer 175

A

Correct Answer: A) Embedding privacy measures throughout the software development life cycle

Explanation: Privacy by Design (PbD) integrates privacy protections from the initial stages of AI system design rather than applying them retroactively.

Answer 176

A

C) Monitor
Explanation: The AI Development Lifecycle consists of four phases: Plan, Design, Develop, and Deploy. Monitoring is part of the ‘Adapt and Govern’ phase post-deployment.

Answer 177

A

A) AI models must undergo impact assessments and document risks
Explanation: Both proprietary AI and high-risk AI applications must comply with transparency, accountability, and risk assessment obligations to meet legal and ethical standards.

Answer 178

A

Correct Answer: B) AI models designed to provide transparent and understandable decision-making

Explanation: Explainable AI (XAI) enhances trust by ensuring AI decisions are interpretable and justifiable, particularly in high-stakes applications like healthcare and finance.

Answer 179

A

B) To identify policy implications of deploying AI
Explanation: The classification framework helps policymakers assess risks and ethical concerns associated with AI deployment.

Answer 180

A

B) Documentation of AI decision-making processes
Explanation: Both frameworks emphasize the need for detailed documentation to ensure AI accountability, transparency, and risk management.

Answer 181

A

Correct Answer: A) The highest level of privacy settings is applied automatically

Explanation: Privacy by Default ensures that users are protected by default settings that uphold strong privacy protections, reducing the risk of data misuse.

Answer 182

A

B) The EU AI Act regulates GPAI at the EU level, while NIST AI RMF provides voluntary guidance on managing their risks
Explanation: The EU AI Act establishes specific regulatory requirements for foundation models like GPAI, while NIST AI RMF offers adaptable risk assessment approaches.

Answer 183

A

A) Implementing explainability and transparency mechanisms
Explanation: Ensuring AI transparency helps identify and mitigate biased decision-making.

Answer 184

A

A) Training a model across multiple devices without data centralization
Explanation: Federated learning allows multiple devices to train AI models collaboratively while keeping data decentralized.

Answer 185

A

Correct Answer: B) Providing sensory input for AI to interpret environmental data

Explanation: Machine perception allows robots to process sensory data such as vision, sound, and touch, enabling autonomous interaction with their environment.

Answer 186

A

B) There is no way to determine if AI training data contains copyrighted material
Explanation: AI models store distributed representations of data, making it difficult to assess whether copyrighted works have been memorized.

Answer 187

A

Correct Answer: B. AI-driven hiring tools with bias detection capabilities

Explanation: While AI in hiring is considered high-risk, it is not banned. AI systems that manipulate individuals or categorize people based on biometric data in a discriminatory way fall under unacceptable risk.

Answer 188

A

B) Track performance metrics and detect anomalies
Explanation: AI dashboards allow real-time monitoring to ensure model stability and detect performance issues.

Answer 189

A

D) The Reformed Product Liability Directive
Explanation: The Reformed PLD applies strict liability to AI, meaning developers and deployers are automatically responsible if an AI product is found defective, without requiring proof of negligence.

Answer 190

A

D) The Reformed Product Liability Directive
Explanation: The Reformed PLD applies strict liability to AI, meaning developers and deployers are automatically responsible if an AI product is found defective, without requiring proof of negligence.

Answer 191

A

A) Both frameworks emphasize the need for ongoing monitoring of AI model performance and risk factors
Explanation: Continuous monitoring ensures that AI models remain compliant, fair, and free of unintended risks over time.

Answer 192

A

Correct Answer: A) Purpose Limitation

Explanation: Purpose Limitation ensures that organizations clearly define and adhere to the intended purpose of collected personal data.

Answer 193

A

B) Identify AI system vulnerabilities through simulated attacks
Explanation: Red teaming tests AI security by simulating real-world attacks to identify weaknesses.

Answer 194

A

A) Data governance ensures that AI models comply with fairness, transparency, and privacy regulations
Explanation: Effective data governance is essential for ensuring compliance with legal, ethical, and operational requirements for AI models.

Answer 195

A

A) The EU AI Act assigns AI systems to specific risk categories, while NIST AI RMF allows organizations to define their own risk management approach
Explanation: The EU AI Act mandates compliance based on predefined risk levels, while NIST AI RMF promotes a flexible, context-based risk management strategy.

Answer 196

A

A) The NIST AI RMF encourages voluntary risk management, while the EU AI Act mandates specific risk mitigation steps
Explanation: While NIST AI RMF provides flexible guidance, the EU AI Act legally requires compliance with structured risk mitigation obligations for high-risk AI models.

Answer 197

A

B) Using explainability tools and conducting bias audits
Explanation: Transparency involves making AI decision-making processes interpretable and conducting audits to detect biases.

Answer 198

A

B) The Equal Employment Opportunity Commission (EEOC)
Explanation: The EEOC enforces laws prohibiting workplace discrimination and has issued guidance on AI-based hiring tools to ensure compliance with Title VII of the Civil Rights Act.

Answer 199

A

Correct Answer: A) Using diverse and representative training datasets

Explanation: Ensuring diverse datasets helps reduce bias, improving AI model fairness and preventing discriminatory outcomes.

Answer 200

A

C) How many hidden layers should the model have?
Explanation: AI planning focuses on business objectives, KPIs, and efficiency rather than specific technical model configurations.

Answer 201

A

Correct Answer: A) Making AI decision-making processes explainable and understandable

Explanation: Algorithmic transparency ensures that AI decisions are interpretable, allowing stakeholders to understand how AI systems generate outcomes.

Answer 202

A

Correct Answer: B) Human-in-the-loop oversight

Explanation: Human-in-the-loop AI requires human intervention in AI decision-making, ensuring safety and ethical alignment in critical applications like healthcare and law enforcement.

Answer 203

A

Correct Answer: A. To ensure responsible AI development and deployment

Explanation: AI governance establishes ethical, legal, and operational frameworks to guide AI development and mitigate risks.

Answer 204

A

B) It is controlled by a private entity with restricted access
Explanation: Proprietary AI models are owned and controlled by private companies, restricting access to their architecture and training data.

Answer 205

A

B) The NIST AI RMF recommends fairness evaluations, while proprietary AI models may lack transparency, making fairness assessments difficult
Explanation: Proprietary models often do not disclose their data sources, making fairness audits harder, while NIST AI RMF recommends organizations implement fairness evaluation practices.

Answer 206

A

Correct Answer: A) AI-powered loan approval systems

Explanation: AI-driven credit scoring and loan approvals impact financial decision-making, requiring fairness, transparency, and risk assessments under the EU AI Act.

Answer 207

A

A) Conduct technical documentation and risk management
Explanation: AI providers must maintain records, assess risks, and document system capabilities.

Answer 208

A

Correct Answer: A) Providing clear justifications for an AI-generated credit approval decision

Explanation: AI explainability ensures that users and regulators can understand how AI decisions are made, improving trust and accountability.

Answer 209

A

Correct Answer: B) Anomaly detection algorithms

Explanation: Anomaly detection identifies patterns that deviate from normal behavior, making it effective for fraud detection in financial transactions.

Answer 210

A

Correct Answer: A) AI governance refers to regulatory compliance, while AI ethics focuses on moral AI principles

Explanation: AI governance ensures compliance with laws and industry standards, while AI ethics addresses broader societal and moral concerns.

Answer 211

A

Correct Answer: B. Severity of harm * probability of occurrence

Explanation: AI risk is typically calculated by assessing the severity of harm an AI system may cause and the likelihood of such harm occurring.

Answer 212

A

Correct Answer: B. To categorize AI systems based on their potential risks and apply appropriate regulations

Explanation: The EU AI Act classifies AI systems into unacceptable, high, limited, and minimal risk categories, imposing stricter requirements on higher-risk applications.

Answer 213

A

A) The requirement for AI providers to conduct bias audits and transparency assessments
Explanation: Both frameworks emphasize the need for AI systems to be explainable and trustworthy, particularly in high-risk applications.

Answer 214

A

Correct Answer: A) GDPR

Explanation: The General Data Protection Regulation (GDPR) enforces strict data privacy protections, requiring AI systems to comply with transparency and user rights.

Answer 215

A

Correct Answer: B) Difficulty in evaluating model performance due to the lack of predefined categories

Explanation: Unlike supervised learning, unsupervised learning lacks labeled data, making it challenging to assess model accuracy and effectiveness.

Answer 216

A

A) Explainability and documentation
Explanation: Third-party AI vendors must provide detailed documentation and explainability to ensure compliance and reduce risks.

Answer 217

A

Correct Answer: B. AI’s ability to function independently

Explanation: AI risk assessment focuses on identifying and mitigating potential harms, considering business use, data integrity, and societal implications.

Answer 218

A

Correct Answer: A) Allows individuals to request deletion of their personal data

Explanation: The Right to Erasure ensures individuals can request the removal of their personal data if it is no longer necessary or processed unlawfully.

Answer 219

A

C) Unacceptable Risk
Explanation: AI systems deemed Unacceptable Risk, such as social scoring and mass biometric surveillance, are banned under the AI Act.

Answer 220

A

C) AI models may inadvertently create biases in credit decisions
Explanation: AI-based lending tools must comply with the FCRA and ECOA, ensuring fair credit decisions and preventing algorithmic bias.

Answer 221

A

A) The requirement for AI companies to delete models trained on unauthorized data
Explanation: The FTC has ordered companies to delete both datasets and models built on improperly obtained personal data, as seen in cases against Amazon and Cambridge Analytica.

Answer 222

A

C) Model hyperparameter tuning
Explanation: Hyperparameter tuning is part of AI development, whereas readiness assessment focuses on opportunity discovery, governance, and infrastructure needs.

Answer 223

A

A) Plan
Explanation: The ‘Plan’ phase focuses on defining the business problem, mission, data requirements, and governance before moving to system design.

Answer 224

A

C) Open-source AI models with high risk potential
Explanation: Open-source AI models are exempt unless classified as high-risk, in which case they must comply with AI Act provisions.

Answer 225

A

B) Protecting users from harmful AI outcomes
Explanation: The AI Bill of Rights is designed to safeguard privacy, prevent bias, and ensure transparency in AI applications.

Answer 226

A

B) ISO/IEC 22989:2022
Explanation: This standard ensures consistent AI terminology to aid policymakers, developers, and auditors.

Answer 227

A

B) Both frameworks require organizations to define AI risks and implement governance policies, which can be complex across different jurisdictions
Explanation: Both frameworks emphasize governance and accountability, requiring organizations to define risk management approaches that comply with different regulatory standards.

Answer 228

A

B) Compliance with industry laws and ethical guidelines
Explanation: AI models in regulated industries must comply with privacy, security, and ethical standards such as GDPR and HIPAA.

Answer 229

A

Correct Answer: A. High-risk AI requires more rigorous compliance measures due to its potential impact on rights and safety

Explanation: High-risk AI must comply with transparency, risk mitigation, and documentation requirements, while GPAI has specific transparency and accountability measures but not as stringent as high-risk AI.

Answer 230

A

A) GPAI models are required to be fully explainable under the EU AI Act, aligning with NIST AI RMF’s best practices
Explanation: The EU AI Act mandates transparency for GPAI, while NIST AI RMF suggests explainability as a best practice for AI risk management.

Answer 231

A

B) Log every AI system decision publicly
Explanation: AI providers must ensure transparency but are not required to make all AI decisions public.

Answer 232

A

Correct Answer: A) AI is trained with labeled data

Explanation: Supervised learning involves training models with labeled input-output pairs, allowing AI to map inputs to desired outputs.

Answer 233

A

Correct Answer: A) Ensuring fairness across diverse demographic groups

Explanation: AI models can exhibit bias if trained on unrepresentative data, making fairness audits crucial in AI governance.

Answer 234

A

A) GDPR
Explanation: The General Data Protection Regulation (GDPR) mandates explainability for AI models that process personal data.

Answer 235

A

B) ISO/IEC 42001:2023
Explanation: This is the first international standard that provides governance structures for AI risk management and compliance.

Answer 236

A

B) They lack transparency and accountability
Explanation: Since proprietary models do not disclose their architecture or training data, they raise concerns about fairness, accountability, and potential biases.

Answer 237

A

B) 35 million euros or 7% of annual global turnover
Explanation: Severe violations result in heavy fines to ensure compliance with AI governance rules.

Answer 238

A

Correct Answer: A) Allows individuals to transfer personal data between service providers

Explanation: The Right to Data Portability ensures users can obtain and reuse their personal data across different services in a structured, machine-readable format.

Answer 239

A

B) Title VII of the Civil Rights Act
Explanation: This law prohibits employment discrimination, including cases where AI hiring systems create bias.

Answer 240

A

A) Both frameworks require organizations to implement data governance policies to ensure AI fairness and privacy
Explanation: Data governance is essential for mitigating AI risks related to fairness, bias, privacy, and compliance with AI regulations.

Answer 241

A

A) The NIST AI RMF provides flexible risk management guidance, while the EU AI Act classifies AI systems into fixed risk levels
Explanation: The EU AI Act enforces strict risk categories, while the NIST AI RMF allows organizations to tailor risk management approaches based on context.

Answer 242

A

Correct Answer: B. AI must be traceable, explainable, and accountable

Explanation: NIST AI RMF emphasizes trustworthy AI by ensuring AI decisions can be traced, explained, and held accountable to human oversight.

Answer 243

A

B) Lack of interpretability
Explanation: Black-box models, such as deep neural networks, lack transparency, making it difficult to understand their decision-making process.

Answer 244

A

D) Plan
Explanation: The AI development lifecycle consists of four stages—Plan, Design, Develop, and Deploy. The planning phase involves defining the business problem, mission, data requirements, and governance.

Answer 245

A

B) Statistical bias in AI training data
Explanation: The NIST AI RMF provides guidelines for detecting and mitigating bias introduced through AI training data.

Answer 246

A

A) Both require organizations to establish clear accountability structures and risk assessment processes
Explanation: The ‘Govern’ function in NIST AI RMF and the EU AI Act’s AI provider obligations both emphasize structured governance and compliance frameworks.

Answer 247

A

B) AI operates autonomously, making it difficult to attribute harm to a specific party.
Explanation: Because AI systems function independently, determining responsibility—whether it’s the developer, the user, or the AI itself—is legally complex.

Answer 248

A

Correct Answer: B) AI bias can lead to unfair and discriminatory outcomes

Explanation: AI bias can result in discriminatory hiring, lending, and policing practices, making fairness and bias mitigation crucial in AI governance.

Answer 249

A

C) Edge
Explanation: Edge computing processes data locally on devices, reducing latency and improving privacy but is limited by hardware constraints.

Answer 250

A

D) Scalability
Explanation: The three key considerations when determining AI project scope are Impact, Effort, and Fit. Scalability is a later-stage consideration.

Answer 251

A

Correct Answer: A. AI applications that are strictly banned due to their potential to cause harm

Explanation: Unacceptable risk AI includes systems like social credit scoring, subliminal manipulation, and real-time biometric categorization in public spaces.

Answer 252

A

A) Data governance ensures compliance with fairness, bias mitigation, and privacy regulations
Explanation: Both frameworks highlight the importance of responsible data management to prevent AI bias, privacy violations, and fairness issues.

Answer 253

A

C) Guaranteed fairness and bias elimination
Explanation: Proprietary AI models are often criticized for lack of transparency and potential bias, as their decision-making processes are not publicly accessible.

Answer 254

A

B) Establishing policies and accountability structures for AI risk management
Explanation: The ‘Govern’ function ensures that organizations have policies and structures in place for AI oversight and risk management.

Answer 255

A

A) Lack of technical documentation
Explanation: Organizations must review vendor policies, documentation, and security measures when using third-party AI models.

Answer 256

A

A) An entity that makes AI available in the EU after importation
Explanation: Distributors make AI systems available as a follow-on action after importation.

Answer 257

A

B) On-premise
Explanation: On-premise AI deployment allows greater data control but requires higher upfront costs for infrastructure and maintenance.

Answer 258

A

B) Provide ethical guidelines for AI design
Explanation: IEEE 7000-21 integrates ethics into AI development, ensuring that systems align with societal values.

Answer 259

A

C) Ensure compliance, fairness, and accountability
Explanation: AI governance post-deployment focuses on compliance, monitoring bias, and ensuring ethical AI operations.

Answer 260

A

A) They must undergo rigorous compliance measures before deployment
Explanation: High-risk AI systems must meet strict requirements for transparency, testing, and oversight.

Answer 261

A

B) Formatting raw data into structured data
Explanation: Data wrangling (or preparation) involves cleaning, structuring, and transforming raw data into a usable format.

Answer 262

A

C) Providing guidance on risk management for generative AI
Explanation: The Generative AI Profile helps organizations identify and mitigate risks associated with AI-generated content.

Answer 263

A

Correct Answer: A) A tool for assessing privacy risks associated with data processing

Explanation: A DPIA identifies, evaluates, and mitigates potential privacy risks before deploying AI systems that handle personal data.

Answer 264

A

A) Proprietary AI models may lack sufficient documentation to meet NIST AI RMF risk measurement standards
Explanation: The limited access to proprietary AI model documentation creates challenges in assessing risks using NIST AI RMF methodologies.

Answer 265

A

Correct Answer: A) Allowing an AI model to apply knowledge from one task to another similar task

Explanation: Transfer learning enables models to leverage pre-trained knowledge, improving efficiency when adapting to new tasks with limited data.

Answer 266

A

B) Continuous risk monitoring
Explanation: NIST AI RMF advocates continuous assessment of AI risks throughout its lifecycle rather than static or one-time evaluations.

Answer 267

A

B) NIST
Explanation: The National Institute of Standards and Technology (NIST) created AI RMF to help organizations manage AI risks and compliance.

Answer 268

A

C) A specific explanation of the factors leading to the adverse decision
Explanation: The ECOA requires creditors to explain specific reasons for denying credit applications, even when using AI, to prevent discrimination and ensure transparency.

Answer 269

A

C) A foundation model with broad applicability across tasks
Explanation: GPAI models, also called foundation models, display generality and can be integrated into multiple applications.

Answer 270

A

Correct Answer: C. 100

Explanation: The AIGP exam consists of 100 multiple-choice questions, designed to assess knowledge of AI governance and risk management.

Answer 271

A

B) Transparency obligations for providers and deployers
Explanation: Article 50 requires AI providers to disclose system capabilities, risks, and transparency measures.

Answer 272

A

B) The potential for an AI system to cause unintended harm or impact trustworthiness
Explanation: AI risk refers to unintended consequences, such as bias, security vulnerabilities, and loss of trust in AI systems.

Answer 273

A

B) Proprietary models often lack transparency, making compliance with NIST AI RMF more challenging
Explanation: The NIST AI RMF emphasizes transparency, but proprietary AI models may limit access to information about their decision-making processes.

Answer 274

A

B) Implementing continuous risk monitoring and documentation
Explanation: Both frameworks stress the importance of ongoing AI risk assessments, audits, and documentation to ensure compliance and responsible AI deployment.

Answer 275

A

A) The EU AI Act requires AI explainability for all AI models, while NIST AI RMF only suggests it as a best practice
Explanation: The EU AI Act mandates transparency for Limited and High-Risk AI models, while the NIST AI RMF recommends explainability as a key principle in risk management.

Answer 276

A

B) Operates exclusively in high-risk sectors
Explanation: GPAI models are used across various domains, not just high-risk applications.

Answer 277

A

B) AI-generated works cannot be copyrighted because they lack human authorship.
Explanation: U.S. copyright law requires human authorship. Cases such as Thaler v. Vidal (2023) confirm that AI cannot be named as an inventor, and the U.S. Copyright Office has repeatedly denied AI-generated works copyright protection.

Answer 278

A

Correct Answer: B. Identifying, assessing, and mitigating risks

Explanation: Risk management is a structured approach to assessing potential AI-related harms and implementing controls to mitigate them.

Answer 279

A

Correct Answer: A) By simulating human reasoning through a structured knowledge base

Explanation: Expert systems use inference engines and rule-based logic to mimic human decision-making in specific domains, such as medical diagnosis or tax preparation.

Answer 280

A

Correct Answer: B) It uses feedback in the form of rewards and penalties

Explanation: Reinforcement learning optimizes decision-making by learning from a series of rewards and punishments.

Answer 281

A

Correct Answer: A) General Data Protection Regulation (GDPR)

Explanation: GDPR incorporates Privacy by Design as a core requirement, mandating that organizations implement data protection throughout system development and operations.

Answer 282

A

B) Market Surveillance Authority (MSA)
Explanation: MSAs are national regulatory bodies that enforce AI Act compliance and conduct audits.

Answer 283

A

Correct Answer: B) Organizations should specify why data is collected

Explanation: Purpose Specification requires that organizations define the reasons for collecting personal data and limit its use to those specified purposes.

Answer 284

A

Correct Answer: B. It provides differentiated requirements based on the potential impact of AI systems

Explanation: The EU AI Act applies stricter rules to high-risk AI while imposing fewer obligations on minimal and limited-risk systems, ensuring a balanced approach to regulation.

Answer 285

A

A) Identifying and contextualizing AI risks
Explanation: The ‘Map’ function helps organizations understand and document AI risks within their operational context.

Answer 286

A

A) The need for increased transparency and disclosure of training data
Explanation: Both proprietary AI models and GPAI raise concerns about the lack of transparency in their training data and decision-making processes, impacting compliance with risk management frameworks.

Answer 287

A

A) They are legally binding like Articles and provide technical details
Explanation: Annexes contain technical specifications referenced in Articles and are updated more frequently.

Answer 288

A

Correct Answer: B. Implement responsible AI practices and risk management

Explanation: AIGPs focus on AI ethics, compliance, risk management, and ensuring AI systems align with legal and organizational standards.

Answer 289

A

Correct Answer: B. Biased decision-making

Explanation: Biased AI can lead to unfair outcomes, discrimination, and violations of consumer protection laws, making it a critical governance concern.

Answer 290

A

C) Edge computing
Explanation: Edge computing allows AI models to run on local devices like smartphones, reducing latency and improving privacy.

Answer 291

A

D) Whether the work is classified as a derivative
Explanation: The four factors of fair use are: (1) purpose and character of use, (2) nature of the copyrighted work, (3) amount and substantiality, and (4) market impact. Derivative works are a separate concept in copyright law.

Answer 292

A

B) Robustness, Security, and Safety
Explanation: This principle emphasizes that AI systems should be resilient to attacks, reliable under stress, and secure throughout their lifecycle.

Answer 293

A

B) Importer
Explanation: Importers ensure that AI systems from third countries comply with EU AI Act standards before entering the market.

Answer 294

A

Correct Answer: B. Implementing AI-driven cybersecurity protocols

Explanation: While AIGPs handle AI governance and risk management, cybersecurity is typically managed by security professionals specializing in AI threats.

Answer 295

A

C) To categorize and tag data for supervised learning
Explanation: Labeled data enables supervised learning models to map input to output correctly.

Answer 296

A

B) Providing voluntary guidelines for managing AI risks
Explanation: The NIST AI RMF is a voluntary framework that helps organizations identify, assess, and mitigate AI risks.

Answer 297

A

A) The lack of transparency in both models makes accountability and compliance more difficult
Explanation: Both GPAI and proprietary AI models often restrict access to their training data and internal operations, complicating compliance with EU AI Act requirements.

Answer 298

A

B) Assessing Risks and Impacts of AI
Explanation: NIST’s ARIA framework focuses on risk and impact evaluation of AI models, including security concerns and bias.

Answer 299

A

A) The EU AI Act mandates compliance for AI providers, while NIST AI RMF provides voluntary guidance
Explanation: The EU AI Act enforces legally binding AI governance obligations, while the NIST AI RMF serves as a voluntary risk management framework.

Answer 300

A

B) To identify AI’s business value and data requirements
Explanation: Readiness assessment ensures AI deployment aligns with business objectives, data governance, and compliance requirements.

Answer 301

A

A) NIST AI RMF provides best practices for AI fairness, while the EU AI Act mandates bias audits for high-risk AI systems
Explanation: While NIST AI RMF suggests fairness strategies, the EU AI Act legally requires bias testing and mitigation for high-risk AI applications.

Answer 302

A

C) Limited processing power
Explanation: Edge computing is constrained by device hardware, which limits the complexity of AI models that can be deployed.

Answer 303

A

Correct Answer: B. To assess how AI affects privacy, discrimination risks, and fundamental rights

Explanation: FRIA helps evaluate potential human rights concerns before AI deployment, ensuring compliance with ethical and legal standards.

Answer 304

A

B) Whether training AI models using copyrighted books without permission constitutes infringement
Explanation: The plaintiffs claimed OpenAI used copyrighted books to train its AI models without consent, but the court ruled that training on copyrighted data did not automatically constitute infringement.

Answer 305

A

A) HUDERAF
Explanation: The Human Rights, Democracy, and the Rule of Law Assurance Framework for AI Systems (HUDERAF) was developed to assess AI’s societal impact.

Answer 306

A

B) Automated credit scoring systems
Explanation: AI used in credit scoring, healthcare, and biometric identification is classified as high-risk.

Answer 307

A

Correct Answer: C) Access and Individual Participation

Explanation: This principle ensures data subjects have the right to view and modify their personal data, enhancing transparency and control over their information.

Answer 308

A

B) The AI Bill of Rights
Explanation: The AI Bill of Rights includes specific protections against bias and unfair algorithmic decision-making.

Answer 309

A

Correct Answer: B. To ensure AI development and deployment is safe, transparent, and respects fundamental rights

Explanation: The EU AI Act aims to regulate AI based on risk levels while fostering innovation and protecting fundamental rights.

Answer 310

A

B) Bias testing and mitigation
Explanation: Bias audits help organizations identify and mitigate unfair AI model behaviors.

Answer 311

A

B) Establishing risk governance structures and transparency reports
Explanation: Both frameworks emphasize governance, requiring organizations to define AI risks and document transparency measures.

Answer 312

A

A) Both frameworks require ongoing risk evaluation and model monitoring to prevent AI failures
Explanation: Both frameworks emphasize that AI models should be continuously evaluated to detect biases, risks, and performance shifts.

Answer 313

A

Correct Answer: C. Recitals

Explanation: Recitals provide context and objectives for the legislation but are not legally binding. Articles and annexes, however, carry legal weight and define obligations.

Answer 314

A

Correct Answer: A) Ensuring AI models provide interpretable and transparent decisions

Explanation: Explainable AI (XAI) improves trust and accountability by making AI decision-making understandable to users and regulators.

Answer 315

A

Correct Answer: B) Matching AI behavior to human goals, ethical values, and societal expectations

Explanation: AI alignment ensures that AI systems act in ways that are consistent with human values, safety requirements, and ethical considerations.

Answer 316

A

B) Organizations must establish clear accountability and compliance structures
Explanation: Both proprietary AI governance and the NIST AI RMF stress the need for organizational accountability and structured AI governance policies.

Answer 317

A

B) Predictive hiring assessments
Explanation: AI in hiring decisions can introduce bias and legal risks, making it a high-risk application requiring strict governance.

Answer 318

A

Correct Answer: D) The Turing Test

Explanation: The Turing Test measures AI’s ability to mimic human intelligence but is not an AI governance framework.

Answer 319

A

Correct Answer: A) Supervised learning requires labeled data, while unsupervised learning does not

Explanation: Supervised learning maps input-output relationships using labeled data, whereas unsupervised learning identifies patterns without predefined labels.

Answer 320

A

B) Creating a dedicated AI ethics and governance board
Explanation: AI governance boards help oversee compliance, risk management, and ethical AI implementation.

Answer 321

A

B) The EU AI Act
Explanation: The EU AI Act places strict obligations on proprietary AI providers, especially in high-risk applications.

Answer 322

A

B) Protect AI models against manipulation and attacks
Explanation: Adversarial testing helps identify vulnerabilities where attackers can manipulate AI model outputs.

Answer 323

A

B) Mitigating and responding to AI risks
Explanation: The ‘Manage’ function focuses on taking action to reduce and manage AI-related risks effectively.

Answer 324

A

B) Helps track model changes and rollback if issues arise
Explanation: Versioning allows organizations to maintain model integrity and revert to previous versions if needed.

Answer 325

A

B) Limited transparency in proprietary AI models makes it difficult to assess risks and biases
Explanation: The ‘Measure’ function in the NIST AI RMF focuses on risk evaluation, which is hindered by the closed nature of proprietary AI models.

Answer 326

A

B) RNN (Recurrent Neural Network)
Explanation: RNNs process sequential data and are commonly used in speech recognition, time-series forecasting, and natural language processing.

Answer 327

A

C) Organisation for Economic Co-operation and Development (OECD)
Explanation: The OECD developed this framework to classify AI systems based on factors like economic context, data inputs, and potential risks.

Answer 328

A

C) OECD AI Principles
Explanation: The OECD AI Principles emphasize fairness, transparency, and accountability to promote responsible AI.

Answer 329

A

B) Continuous AI model monitoring and bias audits
Explanation: Both proprietary AI models and AI governance frameworks stress the importance of monitoring AI performance and conducting regular bias audits.

Answer 330

A

A) It recommends AI governance policies that ensure accountability, transparency, and ethical compliance
Explanation: The ‘Govern’ function in the NIST AI RMF provides best practices for structuring AI risk management policies, including accountability and transparency measures.

Answer 331

A

A) Both must comply with transparency, documentation, and risk management obligations
Explanation: The EU AI Act enforces specific compliance requirements for GPAI and proprietary AI models, including transparency and accountability obligations.

Answer 332

A

A) The requirement for AI companies to delete models trained on unauthorized data
Explanation: The FTC has ordered companies to delete both datasets and models built on improperly obtained personal data, as seen in cases against Amazon and Cambridge Analytica.

Answer 333

A

Correct Answer: A) Applying stricter rules to AI systems with higher risks

Explanation: Risk-based AI regulation tailors oversight requirements based on the potential impact of AI applications, imposing stricter measures on high-risk systems.

Answer 334

A

A) Implementing continuous AI monitoring and bias detection
Explanation: Both frameworks emphasize ongoing monitoring and risk assessments to ensure that AI systems remain compliant, fair, and unbiased over time.

Answer 335

A

A) AI can only function correctly if trained on high-quality data
Explanation: Poor-quality data leads to unreliable AI predictions, reinforcing the need for data cleansing and validation.

Answer 336

A

Correct Answer: B) AI decisions require explainability to ensure transparency and fairness

Explanation: Explainability is crucial for understanding AI-driven decisions, ensuring fairness, and mitigating biases.

Answer 337

A

A) Both frameworks emphasize risk management and accountability
Explanation: The NIST AI RMF and EU AI Act share governance principles, requiring organizations to define AI risks, implement monitoring mechanisms, and ensure ethical accountability.

Answer 338

A

A) The EU AI Act classifies AI risks into categories with mandatory obligations, while NIST AI RMF provides a flexible approach to managing AI risks
Explanation: The EU AI Act assigns AI applications to risk levels (Minimal, Limited, High, and Unacceptable), while NIST AI RMF provides adaptable guidelines for organizations to assess and mitigate AI risks.

Answer 339

A

D) They establish enforceable obligations
Explanation: Recitals explain legislative intent but are not legally binding like Articles.

Answer 340

A

Correct Answer: B) Reinforcement learning optimizes actions based on feedback, while supervised learning learns from pre-labeled datasets

Explanation: Reinforcement learning focuses on sequential decision-making with trial-and-error learning, while supervised learning uses labeled data for training.

Answer 341

A

C) Visibility
Explanation: The 5 Vs of big data are Variety, Velocity, Volume, Veracity, and Value.

Answer 342

A

Correct Answer: A. It ensures AI risks align with an organization’s overall risk strategy

Explanation: AI risk management must be integrated with an organization’s broader risk framework to maintain consistency and regulatory compliance.

Answer 343

A

B) Hyperparameter tuning
Explanation: Hyperparameter tuning is an optimization process, not a testing method. AI testing methods include bias testing, robustness testing, and adversarial testing.

Answer 344

A

Correct Answer: B) The OECD Guidelines (1980)

Explanation: The OECD Guidelines established the foundational principles of Fair Information Practices, which have influenced various privacy regulations globally.

Answer 345

A

B) Silverman v. OpenAI
Explanation: This case involved allegations that OpenAI used copyrighted books to train its AI models without consent, raising questions about AI training data and fair use.

Answer 346

A

B) Encouraging unrestricted AI experimentation
Explanation: The Act ensures AI safety while setting legal boundaries on unacceptable practices.

Answer 347

A

Correct Answer: C. High-risk AI systems

Explanation: High-risk AI systems must comply with strict requirements such as risk management, data governance, human oversight, and transparency to minimize risks to safety and fundamental rights.