Article: We Built Judge. AI. And You Should Buy It

Jack Kieffaber (Harvard Law School), Kimo Gandall (Harvard Law School) and Kenny McLaren (Harvard Law School) recently published, We Built Judge. AI. And You Should Buy It, 2025. Provided below is an Abstract: 

We just built Corporate America’s dream and Asimov’s nightmare—a machine that automates all dispute resolution. Meet Arbitrus; it’s Judge.ai for private arbitration, and we’re going to tell you how it works.

Predictable, efficient, and cheap—very cheap—Arbitrus.ai performs just as well as human arbitrators right now, reducing the cost of dispute resolution from $100k to a flat $10k. And with help from laissez faire tech regulation and the permissive Federal Arbitration Act, it sets the table for the Arbitration State—a looming system of mass private adjudication that will gradually outcompete standard public judiciaries and decrease the relevance of centralized governance.

In Part I, we deliver a primer on our theory of automated judging. In Part II, we present a technical outline of Arbitrus.ai, a full-stack end-to-end AI arbitration service. In Part III, we then apply Arbitrus.ai to 100 hypothetical scenarios, and grade the opinion of the AI arbitrator. We find zero (0) hallucinations; zero (0) incomplete responses; and only two (2) ungrounded sources. In Part IV, we address positive and normative objections to Arbitrus’ efficacy and ethics. And finally, in Part V, we talk about consequences—and the steps between our machine and total automated governance.

There are fewer steps than you’d think; join us as we tip over the first domino.