References: AI governance
Primary source
Section titled “Primary source”Dan Hendrycks. Introduction to AI Safety, Ethics, and Society. Taylor & Francis, 2024. Center for AI Safety, free to read at aisafetybook.com. L9 draws from Chapter 8 (Governance), with the four-layer taxonomy organized around sections 8.4 (Corporate Governance), 8.5 (National Governance), 8.6 (International Governance), and 8.7 (Compute Governance). Sections 8.2 (Growth) and 8.3 (Distribution) inform the broader economic framing the lesson does not work in detail.
| Chapter section | Topic | URL |
|---|---|---|
| Ch 8.4 | Corporate Governance | aisafetybook.com/textbook/corporate-governance |
| Ch 8.5 | National Governance | aisafetybook.com/textbook/national-governance |
| Ch 8.6 | International Governance | aisafetybook.com/textbook/international-governance |
| Ch 8.7 | Compute Governance | aisafetybook.com/textbook/compute-governance |
Verbatim quotes used in the lesson
Section titled “Verbatim quotes used in the lesson”A1 discipline preserved: verbatim from cited sections, no paraphrasing inside quote marks.
- §8.6 International Governance, core framing: “international cooperation is important in order to manage risks from AI.”
- §8.6 International Governance, on nuclear-precedent parallel: “offense-dominant” (used by the chapter to describe both AI and nuclear weapons in the context of agreement-deviation risk).
- §8.6 International Governance, on aviation-certification analogy: “domestic regulators must have certain verification procedures” (used in the context of nations being incentivized toward compliance through market-based access consequences).
- §8.7 Compute Governance, on the lever: “Compute is indispensable for developing and deploying AIs. Restricting access to compute allows control over what AIs are created and used.”
- §8.7 Compute Governance, on tractability: “Compute is physical, excludable, and quantifiable which allows it to be tracked, restricted, and measured.”
The four-layer taxonomy itself (corporate, national, international, compute) is the chapter’s organizing structure; the specific mechanisms named within each layer (RSPs, model cards, EU AI Act provisions, IAEA-style organizations, FLOP-based thresholds) are drawn from the broader AI-policy literature that Ch 8 reflects.
Posture and license
Section titled “Posture and license”Same posture as L1 through L8: the CAIS textbook is © 2026 Center for AI Safety, published by Taylor & Francis, free to read online with no explicit Creative Commons or reuse license. This lesson is a structural mirror with verbatim quotes anchored to specific chapter sections within fair-use limits, link-out only, no embed, no derivative runs.
Suggested companion reading
Section titled “Suggested companion reading”Not required for L9; these are the major works in current AI-policy and compute-governance discussion.
- EU AI Act: the full text is at eur-lex.europa.eu. The Future of Life Institute maintains an accessible structured summary at artificialintelligenceact.eu. Worth reading for the specific instruments (the high-risk categorization, the general-purpose AI obligations, the systemic-risk threshold) the lesson references.
- US Executive Order on AI (2023): the original EO 14110 text is at whitehouse.gov (note: subsequent administrations have revised or rescinded portions; check current status). For policy-context analysis, the Stanford HAI explainer is a good entry point at hai.stanford.edu.
- NIST AI Risk Management Framework: the official document at airc.nist.gov/AI_RMF_Knowledge_Base. The voluntary framework that becomes de facto binding through federal procurement and increasingly through state-level adoption.
- Compute governance: Markus Anderljung, Lennart Heim, et al., “Computing Power and the Governance of AI” (Governance of AI Institute working paper, 2023), at arxiv.org/abs/2402.08797. The most-cited treatment of compute as a governance lever; the lesson’s discussion of compute’s physical-excludable-quantifiable properties is anchored here.
- Responsible Scaling Policies: Anthropic’s RSP at anthropic.com/news/anthropics-responsible-scaling-policy and METR’s evaluation of RSPs as a governance mechanism at metr.org. Worth reading both to see corporate-layer commitments in their current form and the external-evaluator perspective on how well they bind.
- Nuclear-AI analogy: Allan Dafoe et al., “International AI Governance: Lessons from the Past, Frameworks for the Future” (Governance of AI Institute, 2023). The most-extended treatment of what AI governance can and cannot inherit from nuclear non-proliferation, climate, and biosecurity precedents. Available via the Future of Humanity Institute publications archive.
- Bletchley Declaration (2023) and successor AI Safety Summit communiques: the declaration text is at gov.uk. Worth reading as a primary-source example of a multilateral norm-development instrument (international governance without binding-treaty teeth).
What this references file closes
Section titled “What this references file closes”This is the last references file in Track 23. The references files across L1-L9 cumulatively cite roughly 50 papers, frameworks, and policy documents from the AI-safety, formal-fairness, complex-systems, game-theory, machine-ethics, and AI-policy literatures. The track is a structural mirror of Hendrycks’ textbook; the companion reading suggestions are the broader literature the chapter draws on. The reader who works through all nine lessons + their companion reading has the structural-mirror foundation for engaging seriously with the field as it stands.