2.R Review and references
In this lesson you will:
- Recap the main points of Domain 2
- Test yourself on ten exam-style review questions
- Look up the sources cited in this domain's sections
This page closes Domain 2, AI Governance, Ethics, Risk and Compliance. Citation numbers in square brackets in the sections, such as [3], refer to the reference list at the end of this page.
Domain summary
- AI governance extends IT governance to cover data-driven, probabilistic systems that affect people's rights and spread without procurement.
- The EU AI Act regulates by risk tier and role. As amended by Regulation (EU) 2026/1744, Annex III high-risk obligations apply from 2 December 2027 and those for AI in regulated products from 2 August 2028; prohibitions and AI literacy have applied since February 2025.
- NIST AI RMF (Govern, Map, Measure, Manage) is a voluntary risk framework; ISO/IEC 42001 is a certifiable AI management system.
- Policies state intent; standards, baselines and procedures implement it; guidelines advise. An accurate AI inventory is the foundation of everything else.
- Responsible AI principles become real only as requirements, controls and evidence. Fairness has several definitions that must be chosen deliberately and then measured.
- Detailed explanations help people but also help attackers; tailor them.
- Under the ISC2 Code of Ethics, protecting society comes before service to your employer.
- Use quantitative (SLE, ARO, ALE) and qualitative analysis together; tier use cases by impact and autonomy; accept residual risk formally through the right owner.
- Personal data in AI spreads to embeddings, logs, models and vendors; public data is still regulated; record lineage so deletion can follow the data.
- Vendor contracts must restrict data use, require model change notice and give incident notification fast enough for your own deadlines (GDPR 72 hours, NIS2 24-hour early warning, CERT-In 6 hours).
- An AIBOM records models, datasets, software and configuration so you can respond quickly to supply-chain incidents. Responsibility is shared with providers, but accountability is not.
- Oversight needs named authority and a tested stop mechanism outside the model; shadow AI is managed best by offering safe alternatives.
- AI literacy is both a control and a legal duty. Train by role and measure behavior and results, not just completion.
Review questions
Answer all ten before opening the answers.
1. An Indian software company sells an AI tool that ranks job applicants to an employer in France. Under the EU AI Act, what is the company's role, and what is the system's likely tier?
- A. Deployer; limited risk
- B. Importer; minimal risk
- C. Provider; high risk
- D. The Act does not apply outside the EU
2. A board wants an AI management system that an accredited body can certify. Which standard should the CISO recommend?
- A. NIST AI RMF
- B. ISO/IEC 42001
- C. ISO/IEC 23894
- D. OWASP Top 10 for LLM Applications
3. A hiring model shortlists 40% of male applicants and 28% of female applicants. Using the four-fifths rule, what is the impact ratio and conclusion?
- A. 0.70; possible adverse impact, investigate
- B. 0.70; no concern
- C. 1.43; no concern
- D. 0.12; possible adverse impact
4. An AI-enabled fraud incident is expected to cost 5,000,000 rupees each time and happen twice a year. A control costing 2,000,000 rupees a year would reduce the rate to once every two years. What is the annual value of the control?
- A. 2,500,000
- B. 7,500,000
- C. 8,000,000
- D. 5,500,000
5. A business unit wants to deploy a high-risk AI system with a known residual risk. Who should formally accept that risk?
- A. The CISO
- B. The data scientist who built the model
- C. The accountable business owner, within appetite set by the board
- D. The vendor
6. A vendor contract for an AI customer-service platform allows the vendor to notify security incidents within 7 days. Why is this a problem for an EU bank?
- A. It is not a problem if the vendor is ISO/IEC 27001 certified
- B. It may prevent the bank meeting its own short deadlines, such as the GDPR's 72 hours
- C. Incident notification is the bank's duty only
- D. AI vendors are exempt from incident reporting
7. An organization discovers widespread use of public AI chatbots by staff. What is the BEST first response?
- A. Provide an approved enterprise alternative, publish an acceptable use policy and bring uses into the inventory
- B. Block all AI domains immediately
- C. Discipline the staff involved
- D. Ignore it as low risk
8. Which artifact BEST tells an auditor a model's intended use, out-of-scope uses and performance across groups?
- A. Datasheet for the dataset
- B. Exception record
- C. Network diagram
- D. Model card
9. A provider of a credit-scoring model offers customers the model's full confidence scores and feature weights for every decision. What security risk does this MOST increase?
- A. Denial of service
- B. Data drift
- C. Model extraction and membership inference
- D. Shadow AI
10. Your manager asks you to sign off a model you have found discriminates unlawfully against a protected group, because the launch date cannot move. According to the ISC2 Code of Ethics, what should you do?
- A. Sign off, since service to your principal comes first
- B. Refuse to sign off and escalate through proper channels
- C. Post the findings on social media
- D. Quietly retrain the model yourself
Answers and explanations
| Q | Answer | Explanation |
|---|---|---|
| 1 | C | It places the system on the EU market under its own name, so it is the provider; recruitment is an Annex III high-risk use. |
| 2 | B | ISO/IEC 42001 is the certifiable AI management system standard. |
| 3 | A | 28 / 40 = 0.70, below 0.80, so there is possible adverse impact to investigate. |
| 4 | D | ALE before = 5,000,000 × 2 = 10,000,000. ALE after = 5,000,000 × 0.5 = 2,500,000. Value = 10,000,000 − 2,500,000 − 2,000,000 = 5,500,000. |
| 5 | C | Risk acceptance belongs to the business owner with appropriate authority; security advises. |
| 6 | B | A slow vendor clause can make the bank miss GDPR, NIS2 or other deadlines. |
| 7 | A | Safe alternatives plus policy work better than bans that drive use underground. |
| 8 | D | A model card documents intended use, limitations and disaggregated performance. |
| 9 | C | Detailed scores and weights help attackers reconstruct the model and infer training membership. |
| 10 | B | Canons I and II (protect society; act legally and justly) outrank Canon III; escalate properly rather than leak. |
References
Numbers in square brackets in the text refer to this list. Items [1] to [9] are on ISC2's suggested reference list for the AI Security pilot exam. Legal dates reflect the position at 12 November 2026.
ISC2 suggested references
- Sayles, J. (2024). Principles of AI Governance and Model Risk Management: Master the Techniques for Ethical and Transparent AI Systems. Apress.
- Wendt, D. W. (2025). AI Strategy and Security: A Roadmap for Secure, Responsible, and Resilient AI Adoption. Apress.
- Manoj, R. K., Senthilnathan, S., Arunmozhi Selvi, S., Ananth Kumar, T., and Balamurugan, S. (2026). AI Trust, Risk, and Security Management. Wiley-Scrivener.
- Santos, O., and Radanliev, P. (2024). Beyond the Algorithm: AI, Security, Privacy, and Ethics. Addison-Wesley Professional.
- Johnson, R., and Easttom, C. (2020). Security Policies and Implementation Issues, 3rd ed. Jones & Bartlett Learning.
- Crossley, C. (2024). Software Supply Chain Security. O'Reilly Media.
- Lin, B. (2026). Privacy and Security for Large Language Models. O'Reilly Media.
- Buckwell, M., Van daele, S., and Horst, C. (2024). Security Architecture for Hybrid Cloud. O'Reilly Media.
- ISC2 (2026). ISC2 Code of Ethics and ISC2 Code of Professional Conduct. isc2.org/ethics
Frameworks, standards and law
- National Institute of Standards and Technology (2023). AI Risk Management Framework (AI RMF 1.0), NIST AI 100-1. doi.org/10.6028/NIST.AI.100-1
- National Institute of Standards and Technology (2024). AI RMF: Generative Artificial Intelligence Profile, NIST AI 600-1. doi.org/10.6028/NIST.AI.600-1
- ISO/IEC 42001:2023. Information technology — Artificial intelligence — Management system. iso.org/standard/81230.html
- ISO/IEC 23894:2023. Information technology — Artificial intelligence — Guidance on risk management.
- Regulation (EU) 2024/1689 (Artificial Intelligence Act). Official Journal of the European Union. eur-lex.europa.eu/eli/reg/2024/1689/oj
- Regulation (EU) 2026/1744 (Digital Omnibus on AI), amending Regulation (EU) 2024/1689; published 24 July 2026, in force 27 July 2026. Summaries used: Technology.org, 13 October 2026; Cooley, Digital AI Omnibus. Confirm against the Official Journal text.
- Regulation (EU) 2016/679 (General Data Protection Regulation). eur-lex.europa.eu/eli/reg/2016/679/oj
- Government of India. Digital Personal Data Protection Act, 2023 and Digital Personal Data Protection Rules, 2025. Ministry of Electronics and Information Technology. meity.gov.in
- Directive (EU) 2022/2555 (NIS2 Directive), Article 23. eur-lex.europa.eu/eli/dir/2022/2555/oj
- OECD (2019, updated 2024). Recommendation of the Council on Artificial Intelligence (OECD AI Principles). oecd.ai/en/ai-principles
- UNESCO (2021). Recommendation on the Ethics of Artificial Intelligence.
- US Equal Employment Opportunity Commission et al. (1978). Uniform Guidelines on Employee Selection Procedures, 29 CFR § 1607.4(D) (the "four-fifths rule").
- Kirkpatrick, D. L., and Kirkpatrick, J. D. (2006). Evaluating Training Programs: The Four Levels, 3rd ed. Berrett-Koehler.
- Mitchell, M., et al. (2019). Model cards for model reporting. Proceedings of the Conference on Fairness, Accountability, and Transparency. arXiv:1810.03993
- Gebru, T., et al. (2021). Datasheets for datasets. Communications of the ACM 64(12). arXiv:1803.09010
- OWASP CycloneDX. Machine Learning Bill of Materials (ML-BOM). cyclonedx.org
- ISO/IEC 42005:2025. Information technology — Artificial intelligence — AI system impact assessment.
- Directive (EU) 2019/790 on copyright and related rights in the Digital Single Market, Articles 3 and 4. eur-lex.europa.eu/eli/dir/2019/790/oj
- Hardt, M., Price, E., and Srebro, N. (2016). Equality of opportunity in supervised learning. Advances in Neural Information Processing Systems 29. arXiv:1610.02413
- CERT-In (2022). Directions under sub-section (6) of section 70B of the Information Technology Act, 2000, 28 April 2022. cert-in.org.in