The Case of AI vs. The Law
Lead Researcher: Noor Omer
This brief is filed on behalf of no party and against no party. It is filed on behalf of the public, who did not ask to be a party to this case and yet are bound by its outcome regardless. The question before us is not whether artificial intelligence will touch the law. That question has already been answered, and the answer is yes. The question is what kind of law it leaves behind, and whether the rest of us had any say in it.
Whether the integration of artificial intelligence into legal practice and judicial decision-making, left unexamined, will alter not only how law is practiced but who it serves, and whether the consequences of that alteration will remain confined to the legal profession or spread to every person who lives under a legal system, regardless of where that person resides.
STATEMENT OF FACTS
Two changes are underway, and they must be treated separately, because they carry very different weight.
The first concerns the practice of law. Lawyers now draft, summarize, and research with the assistance of artificial intelligence, and this practice has moved well past the pilot stage into daily use. Firms that decline to adopt these tools report falling behind firms that do. The consequence that receives the least attention is the quiet closing of a door. Junior lawyers historically built their judgment by performing the very tasks AI now performs faster. That apprenticeship is narrowing, and no institution has yet proposed what should replace it.
The second concerns the administration of justice itself. Courts in Brazil use artificial intelligence to screen appeals for admissibility. Courts in Colombia use it to analyze cases touching on fundamental rights. Courts in Argentina use it to draft judicial opinions from guided prompts. What began, in each of these instances, as an administrative convenience has in several cases evolved into something with genuine influence over which cases are heard and how they are resolved. A tool that decides which disputes reach a judge is no longer merely assisting the judiciary. It is participating in it.
Every legal system depends on a resource it cannot manufacture and cannot purchase, which is the belief, held by the people who live under it, that the system is fair.
The Effects of Judicial AI Do Not Remain Inside the Courtroom
It is well established that courts do not decide cases in isolation. Each ruling becomes material for the next. This is the ordinary operation of precedent, and it is what makes an error introduced through an AI-assisted decision so consequential. The error does not remain a single mistake, correctable on appeal and then forgotten. It becomes foundation. Future rulings are built upon it, often without anyone examining whether the foundation was sound to begin with.
The same logic governs regulation more broadly. Standards developed for how courts admit AI-generated evidence rarely stay confined to the courtroom. They migrate, almost as a matter of course, into how immigration agencies process claims, how tax authorities flag irregularities, and how private companies govern their own use of AI. A rule written carefully in one courtroom becomes a template used everywhere. A rule written carelessly does the same.
Judicial Legitimacy Is a Shared Resource
Nations extend trust to one another's legal systems as a precondition of trade, investment, and diplomacy. That trust rests on a belief that the other system is fair, and that belief is more fragile than it appears. A judiciary perceived as governed by opaque, unaccountable AI risks that belief directly, and the consequences of that risk are not contained to the jurisdiction in question. They surface in how foreign companies assess the risk of doing business there, and in how other governments calibrate their own dealings with that state.
The Distribution of Capable AI Tools Threatens to Create Two Systems of Justice
A well-resourced judiciary can commission AI systems that are carefully audited, rigorously tested, and subject to meaningful human oversight. A judiciary without those resources faces a harder choice, and often adopts whatever tool is most affordable, sometimes procured from a vendor with no meaningful understanding of the legal tradition or culture into which the tool is being introduced. Left unaddressed, this divergence does not remain a matter of technical capacity. It becomes a matter of justice itself, where the fairness a person can expect from their legal system begins to correlate with the wealth of the nation in which they happen to live.
Industry research conducted in 2026 found that only a small share of organizations using AI in legal work could point to measurable financial gains from doing so.
AI and the Case for Liability
A court may label a decision "AI-assisted," but a label is not a remedy, and it is worth asking plainly whether the people affected by that decision have any real, usable means of asking how the AI reached its conclusion, or whether the label functions instead as a formality that satisfies procedure without satisfying justice.
Responsibility presents a comparable difficulty. When an AI system contributes to an erroneous outcome, the question of who answers for it, the judge who relied on the tool, the court that procured it, or the vendor who built it, remains unsettled in most jurisdictions, and an unsettled question of liability tends, in practice, to resolve itself into no liability at all.
Access presents a further difficulty. The optimistic account holds that AI will open the courthouse to people who could never previously afford representation. The more sober account holds that AI will simply hand a further advantage to whichever party already holds every other advantage, and that the well-funded litigant will always have access to the more capable tool.
Finally, it is fair to ask whether this wave of adoption is driven by demonstrated results or by a more ordinary pressure to appear current. Industry research conducted in 2026 found that only a small share of organizations using AI in legal work could point to measurable financial gains from doing so. Some portion of what is currently described as transformation may be better described as appearance.
Conclusion
Every legal system depends on a resource it cannot manufacture and cannot purchase, which is the belief, held by the people who live under it, that the system is fair. Artificial intelligence has the capacity to strengthen that belief, by reducing delay, easing access, and applying rules with greater consistency than any single human institution has managed before. It carries an equal capacity to weaken that belief, quietly and by increments, until the erosion is noticed only after it can no longer be easily reversed.
This is not a question for lawyers and judges to resolve. The rules being written today, in courtrooms most people will never enter, will shape how fairly the next generation experiences the law, without regard to where they were born or what language they speak. The case remains open. The verdict is still being written, and it will be written whether or not the public participates in writing it.
Disclaimer: This content was produced with the assistance of Claude, developed by Anthropic. DARA assumes the role of the AI system, in this case Claude, for its conclusions. iNNOV8's lead researcher takes sole responsibility for the input.
Content Type:Informational
