Oxford University Press quietly released the most comprehensive scholarly reference on artificial intelligence governance ever compiled. Forty-nine chapters. Seventy-seven contributors. Over 1,000 pages. 2024 Nobel laureate Daron Acemoglu is in there. Anthropic co-founder Jack Clark is in there. Stanford's Daniel E. Ho is in there. And almost nobody in the practicing bar has cracked it open.
Here is the harder part. Large sections of the book are already out of date. The word “ChatGPT” barely appears. There is no chapter on how lawyers should actually use AI without getting disqualified. And the wave of 2025 sanctions cases that are now rewriting the ethical standard for every attorney in America is nowhere to be found, because most of those cases had not happened yet when the manuscript went to press.
If you are serious about AI in legal practice in 2026, you still need this book on your shelf. Let me explain exactly why, and exactly how to use it without being misled.
What this book actually is
The Oxford Handbook of AI Governance, edited by Justin B. Bullock, Yu-Che Chen, Johannes Himmelreich, Valerie M. Hudson, Anton Korinek, Matthew M. Young, and Baobao Zhang, was published by Oxford University Press in 2024 (ISBN 9780197579329). It is organized into nine sections: Introduction and Overview, Value Foundations of AI Governance, Developing an AI Governance Regulatory Ecosystem, Frameworks and Approaches, Assessment and Implementation, AI Governance from the Ground Up, Economic Dimensions, Domestic Policy Applications, and International Politics.
The contributor list is genuinely extraordinary. Daron Acemoglu on the harms of AI. Allan Dafoe of Google DeepMind on theoretical lenses. Jack Clark on information markets. Daniel E. Ho, along with Kurt Glaze, Gerald K. Ray, and Christine Tsang, on AI adjudication at the Social Security Administration. Seth Lazar on power and AI. Carissa Véliz on privacy. Inioluwa Deborah Raji on the anatomy of AI audits. Charlotte Stix on the European Union. Michael C. Horowitz on the international balance of power. Marie-Therese Png on the Global South. Joanna J. Bryson on transnational digital governance. Toby Shevlane on structured access. Cary Coglianese and Alicia Lai on automated administration.
This is what foundational reference material is supposed to look like. You will not find a denser, better-sourced, or more intellectually serious treatment of AI governance anywhere else in English right now.
What the handbook addresses well
Start with Section II, the value foundations. Kate Vredenburgh on fairness, Véliz on governing privacy, Theodore M. Lechterman on accountability, David Danks on governance via explainability, Lazar on power, and Johannes Himmelreich and Désirée Lim on structural injustice. Read those six chapters back to back and you have a working ethical vocabulary for every AI policy your firm will ever write. When a client asks why your engagement letter now includes a generative AI disclosure, you can point at Véliz. When a judge asks why your firm insists on human review of every AI-assisted filing, you can point at Lechterman.
Section III builds the regulatory architecture. Brian Wm. Higgins's Chapter 15 on legal elements of an AI regulatory permit program is the single best short treatment I have read of what licensing AI systems could actually look like. Anthony Aguirre, Peter B. Reiner, Harry Surden, and Gaia Dempsey's chapter on AI loyalty by design maps neatly onto the fiduciary concepts lawyers already understand. Matthijs M. Maas on aligning AI regulation to sociotechnical change will give you the theoretical scaffolding for any debate on whether your state should regulate AI at all.
Section V is where practitioners should spend the most time. Raji's Chapter 25 on AI audits is foundational. If you are building any kind of compliance program that touches AI, this is your starting point. Roel I. J. Dobbe's Chapter 22 on system safety applies the Leveson-tradition safety engineering worldview to AI. Pair those two with Coglianese and Lai on assessing automated administration, and you have the audit theory you need to defend any firm AI policy.
Section VIII is what I wish every practicing lawyer would read. Glaze, Ho, Ray, and Tsang on AI at the Social Security Administration is gold for administrative lawyers and, quietly, for domestic relations practitioners like me. The same due process concerns the authors raise about automated benefits adjudication apply directly to courts that are starting to use AI case management tools in family law dockets. Stephen Caines on surveillance, Beatriz Botero Arcila on smart cities, Nakul Aggarwal and co-authors on healthcare AI, and Nikita Aggarwal on fintech round out a practice-area tour that lets every lawyer find at least one chapter directly relevant to their book of business.
And Section IX, on international politics, is the only place you can currently get a serious side-by-side treatment of U.S., EU, NATO, and Global South AI governance between two covers. Stix on the EU, Jeffrey Ding on U.S.–China technonationalism, Horowitz on the international balance of power, and Png on Global South stakeholders will make you smarter than ninety-five percent of the panelists at your next bar association AI CLE.
What the handbook leaves out, and why that matters
The manuscript went to press in summer 2024. Here is what has happened since then. Every one of these developments is either missing from the book or treated only in passing.
No meaningful discussion of how lawyers actually use generative AI
The word “ChatGPT” appears in the index a small number of times and mostly in footnotes. There is no chapter on the Model Rules of Professional Conduct applied to generative AI. ABA Formal Opinion 512 was issued on July 29, 2024, which is almost certainly why it is not operationalized anywhere in the book. The single most important ethics document governing how American lawyers must use AI is not in this volume.
The federal regulatory map is 180° wrong
The book assumes the Biden administration's AI governance architecture. Executive Order 14110, the 2023 Safe, Secure, and Trustworthy AI order, is treated as the federal baseline. On January 20, 2025, President Trump rescinded EO 14110. On January 23, 2025, he signed Executive Order 14179, “Removing Barriers to American Leadership in Artificial Intelligence,” 90 Fed. Reg. 8741 (Jan. 31, 2025). On July 23, 2025, the White House released “Winning the Race: America's AI Action Plan” along with three additional AI executive orders. OMB guidance M-24-10 and M-24-18 have been replaced by M-25-21 and M-25-22. None of this is in the book. Any reader who treats the handbook as a current federal regulatory map will be wrong by about 180 degrees.
The 2025 copyright litigation that reshaped AI training-data law
Thomson Reuters Enterprise Centre GmbH v. Ross Intelligence Inc., No. 1:20-cv-613-SB (D. Del. Feb. 11, 2025), held that copying 2,243 Westlaw headnotes to train a competing research product was not fair use. Bartz v. Anthropic, No. 3:24-cv-05417 (N.D. Cal. June 23, 2025), held that training on lawfully acquired books was transformative fair use but that maintaining a central library of pirated works was not, and produced a $1.5 billion settlement preliminarily approved on September 25, 2025. Kadrey v. Meta reached a parallel training-is-fair-use conclusion two days after Bartz. And on January 5, 2026, Judge Sidney Stein affirmed a magistrate judge's order compelling OpenAI to produce twenty million ChatGPT logs in the consolidated In re OpenAI, Inc. Copyright Infringement Litigation MDL in the Southern District of New York. None of this is in the handbook.
The AI hallucination sanctions wave
Mata v. Avianca, Inc., 678 F. Supp. 3d 443 (S.D.N.Y. 2023), is where the phrase entered the legal lexicon. Since then: Park v. Kim, 91 F.4th 610 (2d Cir. 2024); Gauthier v. Goodyear Tire & Rubber Co.; Coomer v. MyPillow, Inc. (D. Colo. July 7, 2025); Johnson v. Dunn, No. 2:21-cv-1701 (N.D. Ala. July 23, 2025), which disqualified a large firm and referred its attorneys to bar authorities in every jurisdiction where they were admitted; and Shelton v. Parkland Health, No. 3:24-cv-2190, 2025 WL 3012828 (N.D. Tex. Nov. 10, 2025), which held that attorney good faith is no defense to Rule 11 sanctions for AI-fabricated citations. Researcher Damien Charlotin's database tracked over 230 hallucination cases worldwide by July 2025, and the Ohio Supreme Court reported over 350 cases by April 2026 (single source — verify against a second independent count before citing). This is the most consequential body of practice-relevant AI law in the United States, and it is entirely absent from this handbook.
The state AI law patchwork
The Colorado AI Act, SB 24-205, was signed on May 17, 2024, and made it into the book only in passing. Governor Polis's August 28, 2025 postponement of the effective date to June 30, 2026 did not. Texas TRAIGA, HB 149, signed June 22, 2025, effective January 1, 2026, did not. California SB 53, the Transparency in Frontier Artificial Intelligence Act, signed September 29, 2025, effective January 1, 2026, did not. Utah SB 149, the New York RAISE Act, and New York City Local Law 144 are all either absent or mentioned only briefly. These are the laws that actually bind lawyers and their clients in 2026.
No coverage of agentic AI or the Model Context Protocol
Anthropic released MCP on November 25, 2024. The shift from chat interfaces to tool-using agents that can execute multi-step legal workflows is the story of 2025 and 2026 in legal technology. It is not in the book.
The operational detail on EU AI Act enforcement is thin
Stix's Chapter 46 is excellent as a political analysis of the Act's drafting, but the phased implementation timeline (Article 5 prohibitions on February 2, 2025; GPAI obligations on August 2, 2025; high-risk systems on August 2, 2026; full compliance on August 2, 2027) and the July 2025 GPAI Code of Practice are not treated at the depth a compliance lawyer needs.
Almost no practical guidance for small and mid-size firms
The book is heavy on federal agency governance and international organization governance. It is light on what a five-attorney Ohio practice should actually do on Monday morning. That is where LegalTek.ai lives.
Contributor demographics tilt heavily toward the Global North
Png's Chapter 48 is a deliberate corrective and an excellent one, but the gap is real. Practicing lawyers, judges, and in-house counsel are thin on the ground among the seventy-seven contributors.
How to actually use this book
Despite the gaps, this is a book every serious AI-aware lawyer should own, and here is exactly how I use it.
Use it as a citation spine
When you write a brief, a firm AI policy, or a CLE paper, this handbook gives you peer-reviewed, authoritative sources for almost every foundational claim. Opposing counsel cites a Medium post. You cite Lechterman, Chapter 8 in Bullock et al., The Oxford Handbook of AI Governance (Oxford Univ. Press 2024). That is a hierarchy a judge notices.
Use Section II as your ethical vocabulary
Map each of your internal policies to fairness, privacy, accountability, explainability, and power. At Mishak Law, our internal framework COUNSEL operationalizes ABA Formal Opinion 512 across seven dimensions, and each dimension cites back into this handbook. That is how you get a firm policy that survives a disciplinary inquiry.
Use Sections IV and V as the architectural blueprint for a governance program
At SilverTung, our G3M framework operationalizes NIST AI RMF. The handbook gives you the theoretical load-bearing walls: Dobbe on system safety, Raji on audits, Coglianese and Lai on automated administration. Pair the theory with NIST RMF 1.0 and the July 26, 2024 Generative AI Profile and you have a defensible program.
Use Section VIII as practice-area training material
Assign chapters to associates by practice group. Administrative lawyers get Glaze, Ho, Ray, and Tsang. Healthcare counsel get Aggarwal and colleagues. Fintech and consumer finance lawyers get Nikita Aggarwal. Municipal lawyers get Botero Arcila. Build a firm-wide reading rotation and treat it like a brown-bag series.
Read it with a 2026 corrective
This is the critical practice. Read each chapter, and then search what has happened in that domain since August 2024. The chapters give you the analytic frame. The last twenty months fill in the facts. Stix on the EU AI Act pairs with the current European Commission AI Office implementation page. Higgins on regulatory permits pairs with California SB 53 and Texas TRAIGA. Ho on SSA adjudication pairs with recent due process litigation involving automated benefits decisions. This is the reading method that makes the book useful rather than misleading.
Use it for thought leadership
Cite specific chapters in LinkedIn posts, law review articles, bar journal pieces, and CLE presentations. When I write about state-level AI licensing, I cite Higgins. When I write about AI in benefits adjudication, I cite Glaze and Ho. That move — citing Oxford University Press rather than a content farm — changes the professional register of your writing instantly.
The bottom line
The Oxford Handbook of AI Governance is not the book you buy to learn what is new. It is the book you buy to learn what is foundational. It gives you vocabulary, frameworks, and citation authority that no other single volume currently offers. What it does not give you is this week's case law, last quarter's executive order, or any usable guidance on how your firm should actually deploy AI in 2026. For that, you need practitioner resources, current ethics opinions, and ongoing intelligence.
That is what LegalTek.ai exists to provide. The handbook is the why. We are the how. If you are building an AI governance program for your firm, advising clients on the state AI law patchwork, or trying to keep your ChatGPT habit from becoming a disciplinary complaint, you need both.
Forty-nine chapters. Seventy-seven contributors. Over 1,000 pages. Your move.
The 2026 corrective: pair the handbook with our practitioner resources
Use the handbook for vocabulary and citation authority. Use these for the case law, ethics opinions, and operational frameworks the manuscript could not include.
Matthew A. Mishak
Managing Attorney of Mishak Law LLC in Amherst, Ohio, and Founder & CEO of LegalTek.ai LLC (d/b/a SilverTung), an AI-powered legal practice management platform built for Ohio domestic relations practitioners. Author of the forthcoming book AI Agents in Legal Practice: The Definitive Guide.
Not legal advice. This article is for informational and educational purposes only and does not constitute legal advice. It does not establish an attorney–client relationship. Consult qualified counsel licensed in your jurisdiction before acting on any matter discussed here. LegalTek.ai is a technology company, not a law firm.
Appendix: Links & Citations
The Handbook itself
ABA Ethics Authority
Federal Executive Action
International Standards & Frameworks
European Union
United States State Laws
Key AI Case Law (selected)
- Mata v. Avianca, Inc., 678 F. Supp. 3d 443 (S.D.N.Y. 2023)
- Park v. Kim, 91 F.4th 610 (2d Cir. 2024)
- Thomson Reuters Enter. Ctr. GmbH v. Ross Intel. Inc., No. 1:20-cv-613-SB (D. Del. Feb. 11, 2025)
- Bartz v. Anthropic, No. 3:24-cv-05417 (N.D. Cal. June 23, 2025)
- Kadrey v. Meta Platforms, Inc., No. 3:23-cv-03417 (N.D. Cal. June 25, 2025)
- In re OpenAI, Inc. Copyright Infringement Litig., MDL No. 3143 (S.D.N.Y.)
- Johnson v. Dunn, No. 2:21-cv-1701 (N.D. Ala. July 23, 2025)
- Shelton v. Parkland Health, No. 3:24-cv-2190, 2025 WL 3012828 (N.D. Tex. Nov. 10, 2025)