David Hoffman is the William A. Schnader Professor of Law at the University of Pennsylvania Carey Law School and one of the most-cited contracts scholars in the United States. His work brings together deep scholarly expertise in issues raised by AI- and Blockchain-mediated contracts with substantial high-stakes dispute resolution experience.
His scholarship spans smart contracts, decentralized finance, form contracts, and AI-assisted contract interpretation, including foundational work in the Columbia Law Review, NYU Law Review, and Minnesota Law Review, as well as a 2026 AAA white paper on legal infrastructure for agentic commerce. He serves as Special Master in MDL-2323 and on the National Opioid Settlement arbitration panels. David Hoffman shares his perspective and experience in managing disputes in the evolving Web3 industry.
Q. Tell us about your professional background and how you became involved in Web3-related disputes, including matters involving blockchain, cryptocurrency, digital assets, or smart contracts.
I came to Web3 through contract theory rather than the other way around. After clerking in the Eastern District of Pennsylvania and litigating at Cravath, I spent twenty years writing about how contracts actually work — how they are formed, read, breached, and remedied — with a particular focus on contract production and the gap between what written agreements say and what parties expect.
That doctrinal lens turned out to be exactly the right one for blockchain. In 2019, my coauthors and I published Coin-Operated Capitalism in the Columbia Law Review, an empirical study of the disconnect between what initial coin offering smart contracts promised investors in their white papers and what the underlying Solidity code actually did. The answer, for a striking number of issuances, was "not the same thing." That project led to Transactional Scripts in Contract Stacks in the Minnesota Law Review, which examined how on-chain code, off-chain agreements, and platform terms compose into a single transaction.
More recently, I have written on AI-assisted contract interpretation in Generative Interpretation (NYU Law Review, 2024) and co-authored a 2026 white paper with AAA’s President and CEO, Bridget McCormack, titled, Agentic Commerce Needs Legal Infrastructure, and the Courts Are Coming, which argues that AI-mediated transactions will generate disputes that the existing system is not yet built to handle. My ADR work — particularly as Special Master in the NFL Concussion MDL since 2020 — has given me extensive experience adjudicating the kinds of contract-interpretation, eligibility, and proof-of-fact questions that Web3 cases tend to surface.
Q: What types of Web3 disputes do you most often see in your work? How do they differ from more traditional commercial, financial, or technology disputes?
The most consequential Web3 disputes I encounter — and the ones I have spent the most time thinking about — are interpretation disputes: cases where the parties disagree about what their agreement actually requires, and where the "agreement" is no longer a single document but a stack of them.
In a typical smart-contract or agentic-commerce matter, you are looking at a written contract, a set of platform terms, the on-chain code that actually executed, and, increasingly, the cryptographically signed mandate an AI agent acted under. Those layers often say different things, and someone has to decide which one governs and what it means.
That is what makes these cases meaningfully different from traditional commercial or technology disputes. In an ordinary breach case, the neutral works mostly with text and trade usage. Here, the neutral has to consider code as a source of meaning, reconcile it with natural-language terms drafted at a different time and for a different audience, and frequently confront performance that has already executed automatically and cannot be unwound.
My academic research on finding meaning in contract stacks and the use of AI to help with interpretation has direct application to Web3 disputes, where the interpretive task is harder than usual, and the evidentiary record is more machine-readable than usual.
The other recurring categories I see — governance and fiduciary disputes inside tokenized ventures, oracle and event-trigger failures, custody and key-control fights, and composability cases where losses propagate through a stack of protocols — share a common feature: they all turn on questions of meaning across heterogeneous sources. That, to me, is the through-line.
Q: What trends are you seeing in Web3 disputes? What developments do you think will most shape the future of dispute resolution in this area?
The trend I am tracking most closely is the convergence of Web3 with agentic commerce, and more importantly, the role generative AI will play in resolving the disputes that result.
On the commerce side, AI agents are beginning to transact under cryptographically signed mandates, and many of those mandates will settle on stablecoin and smart-contract rails because programmable money pairs naturally with programmable buying. That will multiply disputes about whether an agent exceeded its mandate, whether user intent was properly captured, and which counterparty in a multi-stakeholder stack — model provider, agent platform, payment rail, merchant — bears responsibility when something goes wrong.
On the resolution side, I see three developments shaping what comes next.
- The first is the maturation of generative AI as an interpretive tool for neutrals and counsel. Generative Interpretation showed that LLMs can do serious work assessing the meaning of contract language in full context. Subsequent research, including my forthcoming work on AI-assisted reasoning in contract disputes, suggests that AI is particularly well-suited to the layered, code-plus-text, multi-document interpretive problems Web3 cases generate. I expect routine use within a few years, with the open questions being about transparency, validation, and the proper role of human judgment — questions the AAA and others are already engaging seriously.
- The second is institutional: dedicated panels like this one, and a growing community of neutrals comfortable with both code and contract law, will give sophisticated parties confidence that there is somewhere serious to bring these matters.
- The third is doctrinal maturation. The early caseload was dominated by fraud and collapse — exchange failures, rug-pulls, broken token economics — and the threshold legal questions were quite basic. The next wave will be more like regular commercial disagreement: governance fights inside well-functioning DAOs, breach claims under tokenized contracts, and code-update disputes between sophisticated counterparties.