Chapter 32: Exercises — Global AI Governance Frameworks

Comprehension Exercises

Exercise 1: Governance Instrument Identification For each of the following AI governance instruments, identify: (a) the year it was adopted; (b) the body that adopted it; (c) whether it is binding or voluntary; and (d) its primary substantive focus. - OECD AI Principles - UNESCO Recommendation on the Ethics of AI - G7 Hiroshima AI Process Guiding Principles - G7 Code of Conduct for Advanced AI - EU AI Act

Exercise 2: The Jurisdiction Gap — Applied Analysis An AI recruitment screening tool is developed by a startup incorporated in the Cayman Islands. The engineers who built it work from offices in Poland and Vietnam. The model was trained on data from job platforms in the United States, United Kingdom, and Australia. The tool is sold as a software-as-a-service product used by companies in Germany, Brazil, Singapore, and Canada to screen job applicants.

Map the jurisdiction question: Which national regulatory authorities might have jurisdiction to investigate this tool if it produces discriminatory outcomes? What are the practical barriers to enforcement in this case? What international governance mechanisms, if any, could address this gap?

Exercise 3: Soft Law vs. Hard Law Matrix Create a table comparing the OECD AI Principles, the UNESCO AI Recommendation, and the EU AI Act across the following dimensions: legal bindingness, enforcement mechanism, scope of coverage (who must comply), geographic reach, penalty for non-compliance, and monitoring mechanism. What does this comparison reveal about the current state of global AI governance?

Exercise 4: OECD Principles Gap Analysis The five OECD AI Principles call for inclusive growth, human-centred values, transparency, robustness, and accountability. Identify one major AI deployment from the past two years that you believe violated each of these principles, and briefly explain the nature of the violation. For each, identify whether any existing governance mechanism addressed the violation and, if not, what governance mechanism might have prevented it.

Exercise 5: UNESCO Implementation Assessment The UNESCO Recommendation includes eleven areas of action. Select any three countries — one OECD member, one G20 non-OECD member, and one least-developed country — and research whether each has taken concrete implementation steps in each of your three selected action areas. What patterns do you observe? What explains the variation?

Analysis Exercises

Exercise 6: The Three-Bloc Dynamic — Sector Analysis Choose one AI application sector: facial recognition, AI-powered credit scoring, or AI content recommendation. Analyze how each of the three major governance blocs (EU regulation-first, US market-first, China state-control) approaches governance of this application. Where do they agree? Where do they conflict? What does each approach imply for a company that wants to operate globally in this sector?

Exercise 7: The Brussels Effect — Case Application A US-based company operates an AI system that assists doctors in diagnosing cancer from medical imaging. The company has customers in the United States, Canada, the UK, Germany, France, and Japan. Analyze the Brussels Effect implications:

(a) Under the EU AI Act's risk classification, what tier would this system likely fall into? (b) What specific requirements would this classification impose? (c) Would it be more cost-effective for the company to maintain separate EU-compliant and non-EU-compliant versions of the product, or to apply EU standards globally? What factors would determine this decision? (d) What other jurisdictions' requirements might apply simultaneously with the EU AI Act?

Exercise 8: Voluntary Framework Effectiveness — Historical Comparison The chapter notes that soft law in AI governance may follow a pattern similar to the early development of international human rights law. Research the following: How did the Universal Declaration of Human Rights (1948) influence subsequent binding human rights law? What mechanisms translated non-binding norms into enforceable obligations over time? What analogies and disanalogies exist between the UDHR's trajectory and the current trajectory of international AI governance instruments?

Exercise 9: Geopolitical Analysis — G7 vs. G20 AI Governance The G7 Hiroshima AI Process involves seven wealthy democracies. The G20 encompasses these plus China, India, Russia, Brazil, Saudi Arabia, Indonesia, and others. Research the G20's engagement with AI governance and compare it to the G7 Hiroshima Process. What can the G20 achieve that the G7 cannot? What makes G20 AI governance consensus more difficult to achieve? Is the G7 Hiroshima Process a step toward global AI governance or a substitute for it?

Exercise 10: Civil Society Representation Audit Review the membership of any two of the following governance bodies: the GPAI Council of Ministers, the GPAI Multistakeholder Expert Group, the UNESCO AI Ethics Recommendation drafting committee (based on available documentation), and the UN AI Advisory Body.

(a) What proportion of members are from countries in the Global South? (b) What proportion represent civil society organizations (as distinct from government or industry)? (c) What proportion are women? (d) What governance changes would you recommend to improve representation?

Applied and Strategic Exercises

Exercise 11: Multinational Compliance Roadmap Your company is a global financial services firm with operations in 35 countries. You are developing an AI system that will assist loan officers in making lending decisions, incorporating traditional credit bureau data, open banking transaction data, and alternative data including mobile payment history. Draft a one-page compliance roadmap that: - Identifies the major governance frameworks and regulations that apply to this system - Prioritizes compliance requirements in order of legal exposure - Identifies the key documentation, technical, and organizational requirements - Notes areas of significant regulatory uncertainty

Exercise 12: National AI Strategy Comparison Compare the national AI strategies of any three countries — one from the EU, one from the Global South, and one from Asia outside the EU and China. Use the OECD AI Policy Observatory (oecd.ai) to access these documents.

(a) How does each country frame the purpose of AI governance? (b) What governance mechanisms does each propose? (c) How does each balance innovation promotion with rights protection? (d) What does each omit that you think should be included?

Exercise 13: The Race to the Bottom — Empirical Investigation The "race to the bottom" in AI governance is an often-cited concern. Research whether evidence exists that it is actually occurring. Identify at least two cases where you can document that a company or government made AI governance decisions based partly on the desire to exploit regulatory arbitrage opportunities. What was the outcome? Was the race to the bottom "won" — i.e., did the least protective standard prevail — or did other mechanisms (reputational pressure, consumer backlash, regulatory catch-up) intervene?

Exercise 14: G7 Code of Conduct Gap Analysis Review the publicly available text of the G7 Hiroshima AI Process Code of Conduct. Identify three specific commitments in the Code of Conduct and, for each: (a) What specific organizational practices would constitute compliance? (b) How could compliance be verified? (c) What incentives exist for non-compliance? (d) Is there any mechanism to detect or sanction non-compliance?

What does this analysis reveal about the limitations of voluntary codes of conduct as governance mechanisms?

Exercise 15: The OECD AI Policy Observatory Spend 30 minutes exploring the OECD AI Policy Observatory at oecd.ai. Identify: (a) The five countries whose national AI policies most closely align with OECD AI Principles in terms of concrete implementation (b) The five countries whose national AI policies show the largest gap between stated principle endorsement and concrete implementation (c) Two emerging governance approaches that the Observatory has identified that were not covered in this chapter

Critical and Reflective Exercises

Exercise 16: Regulatory Imperialism or Global Public Good? The "Brussels Effect" and the EU AI Act's extraterritorial scope mean that EU regulatory choices are effectively imposed on people in countries that had no democratic input into those choices. Is this: (a) Regulatory imperialism that undermines democratic self-governance of non-EU countries? (b) A legitimate exercise of market power that happens to raise global standards? (c) The best available mechanism for global AI governance given the difficulty of achieving binding international treaties?

Write a 500-word analysis defending one of these positions while engaging seriously with the strongest arguments for the other positions.

Exercise 17: China's AI Governance — What Can Be Learned? China has produced technically sophisticated AI governance regulations — on algorithmic recommendations (2022), deep synthesis (2022), and generative AI (2023) — that in some respects are more detailed than Western equivalents. Yet their purpose is fundamentally different from rights-protective governance frameworks.

(a) What specific provisions in Chinese AI regulation reflect genuine consumer or rights protection, independent of political control objectives? (b) What can Western AI governance frameworks learn from Chinese regulatory approaches, if anything? (c) Is it possible to separate the "technique" of China's regulatory approach from its political purposes?

Exercise 18: UN AI Governance — The ITU Problem Research the International Telecommunication Union's engagement with AI governance and the concern, expressed by some Western governments and civil society organizations, that Chinese actors have used ITU standard-setting processes to legitimize surveillance technology and establish governance norms that favor authoritarian applications.

(a) What specific ITU AI standard-setting processes have raised concerns? (b) Are the concerns well-founded? (c) What are the implications for decisions about whether to route global AI governance through the UN system?

Exercise 19: What Would a Global AI Treaty Require? The chapter argues that binding international AI agreements face formidable obstacles. Engage in reverse engineering: What would a meaningful global AI treaty need to contain, and what political conditions would need to exist for it to be ratified? Specifically:

(a) What substantive commitments would be worth making legally binding at the global level? (b) What verification mechanisms would be needed? (c) What dispute resolution mechanism would be required? (d) What political conditions — in US-China relations, in multilateral institutions, or in civil society — would need to change?

Exercise 20: Global South AI Governance Agenda You are advising a coalition of African, South Asian, and Latin American governments on their priorities for global AI governance processes. What specific reforms to existing governance processes and institutions would you recommend? Draft a one-page position paper that: - Identifies the three most urgent reforms to existing global AI governance processes - Explains why these reforms would benefit Global South nations - Addresses likely objections from wealthy-country governments and major AI companies

Scenario and Simulation Exercises

Exercise 21: Diplomatic Simulation — AI Governance Negotiation Working in groups of five, simulate a diplomatic negotiation over a proposed binding international AI agreement on facial recognition in law enforcement. Assign the following roles: - EU representative (favoring strict prohibitions) - US representative (favoring voluntary standards with law enforcement exceptions) - Chinese representative (opposing prohibitions that would apply to sovereign security decisions) - Civil society representative (favoring strict prohibition without exception) - Representative of a developing-country coalition

Negotiate for 30 minutes and report on: whether an agreement was reached, what its content was, and what compromises were required. Debrief on what the simulation reveals about the difficulty of binding international AI governance.

Exercise 22: Incident Response — Cross-Border AI Harm An AI system developed by a South Korean company and deployed by a UK financial institution has produced systematically discriminatory credit decisions affecting consumers in France, Germany, and Italy. The system uses training data from US credit markets that does not adequately represent European consumer credit patterns.

(a) Which regulatory authorities have jurisdiction, and what authority does each have? (b) What investigative steps would each authority likely take? (c) What remedies are available to affected consumers? (d) What governance mechanism could have prevented this outcome?

Exercise 23: Standards Body Engagement Your organization has been invited to participate in an ISO working group developing AI governance standards. The working group will produce technical specifications for AI risk assessment documentation that will be referenced in procurement contracts globally.

(a) What governance principles from this chapter should be reflected in the standards? (b) What specific provisions would you advocate for? (c) What industry arguments against strong documentation requirements would you anticipate, and how would you respond? (d) How should civil society voices be incorporated into this technical standards process?

Exercise 24: GPAI Research Priorities The Global Partnership on AI is soliciting proposals for new research working group projects. Design a research proposal that addresses an AI governance challenge affecting Global South countries that is currently underaddressed in GPAI's research agenda. Your proposal should: - Identify the specific governance challenge - Explain why existing research is inadequate - Propose a research methodology - Identify who should be involved in the research - Explain how the research findings would improve governance outcomes

Exercise 25: Organizational Governance Commitment Assessment Select a major AI company (Google, Microsoft, OpenAI, Meta, Amazon, or an AI company of your choice) and evaluate the gap between its stated commitments to global AI governance principles and its actual practices.

(a) What specific commitments has the company made to global AI governance frameworks (OECD Principles, G7 Code of Conduct, etc.)? (b) Identify at least two documented cases where the company's AI practices appeared to contradict its stated governance commitments. (c) What accountability mechanisms exist to identify and address this gap? (d) What governance reforms would you recommend to close it?