We have a special excerpt for you from this fascinating new book!
By Bruce Schneier and Nathan E. Sanders, extracted/adapted from Rewiring Democracy
In 2023, much was made of GPT-4’ s supposedly excellent performance on legal proficiency tests like the LSAT and bar exams. Improvement on standardized tests is interesting, but today’s AIs are not yet up to the task of performing the high-stakes, highly detail-oriented, highly cognitively complex work of lawyering. And yet we should root for AI to keep improving here, because access to legal counsel is essential to justice, and the high cost of effective representation remains a barrier to access in many democracies. Perhaps AI assistance could lower the cost of lawyering.
However, the result could be to leave the most vulnerable people with an ineffective AI lawyer. Already, public defenders and legal aid offices are often strapped for funding and struggle to attract attorneys who could earn much more serving wealthier clients. As a case in point, 98% of US federal convictions result from plea bargains, often arising because defendants don’t have the resources or competent representation needed to go to trial. Stripping back access to representation even further by using AI assistance as an excuse to give public defenders a larger caseload, or to replace human public defenders with AI alone, could exacerbate this problem. One defendant tried to use an AI avatar to present his case to a New York court in 2025 (the justices did not allow it).
However, a responsible deployment of AI could save public defenders time and legitimately increase their performance and capacity. In 2023 in the US, the Miami-Dade County’s Public Defender office touted itself as the first in the nation to provide AI tools to its attorneys, with the intent of supporting attorneys’ “wellbeing.” We don’t yet know the outcome of this experiment, but it is an early signal of a real trend towards AI assistance in the legal field.
One surefire way to concentrate power and reinforce inequality is to use tools asymmetrically in the legal system. In the US, AIs are already being used by some defense counsels to aid in selecting jurors. AI systems can ferret out online information about jurors that could be used to disqualify them. AI simulations of judges, opposing counsel, and actual jurors can be used to test different ways to present a case. Selective use of those capabilities would not increase access to justice; it would further concentrate the power of those who can afford the best AIs. But keep in mind that this is already true: well-resourced counsels make extensive use of expensive research services and jury consultants. If AI provides a cheaper route to equivalent capabilities, it could help even the playing field.
The various AI applications could either widen or narrow the gap between the rich and the poor, depending on how well they work. Recall in chapter 21, “Augmenting Versus Replacing People,” when we talked about the two ways AI-assisted expertise could affect a profession. It could primarily raise the average professionals’ performance, or it could primarily enhance the top performers. If highly paid top lawyers suddenly become less valuable because AI-assisted junior lawyers are equally productive, then access to quality lawyering should become more widely affordable. On the other hand, if those top lawyers become even better at their work thanks to AI assistance, then the gap between the quality of legal representation affordable by the wealthy and to those of average means will grow.
Efficiencies enabled by AI can make a difference to all parties in the legal system, including the government’s own representation. In June 2024, Brazil’s solicitor general announced that his office would begin to use AI tools to triage thousands of lawsuits involving the federal government. Since Brazil pays out more than 1% of its GDP annually in court-ordered debts, a small increase in the efficiency of its legal defense could have a big impact on its budget.
AI will change what it means to file and receive a legal complaint, which would affect every party to the legal system. Today, the cost of hiring a lawyer and commencing legal proceedings represents a strong social signal due to the expense involved. If writing a threatening legal letter becomes as easy as prompting a chatbot, then it will no longer send the same social signal. (Such a change should be familiar to readers. Only two decades ago, if someone remembered your birthday, it signaled that they were truly your friend. Now that Facebook automatically reminds everyone of birthdays, it no longer has the same meaning.) If drafting a complaint and filing a lawsuit becomes both easy and cheap, the result may be a huge increase in the number of court filings that would overwhelm the legal system—unless the courts also have AI tools to control the flood.
Paradoxically, increasing the ease of initiating litigation may result in fewer lawsuits. Much like criminal plea bargaining, many civil legal disputes settle rather than go to trial. In the US, UK, Australia, and other nations whose legal system originated in English law, more than 90% of cases conclude by settlement. Attorneys may prefer to settle because they are confident in their ability to estimate their chances of winning at trial and calculating potential settlement payments. To the extent that AI simulations are capable of providing attorneys with even more information, we could see fewer cases filed and earlier settlement of more cases.
The emergence of AI-assisted lawyers will have profound implications for the profession. An influx of AI could draw new types of talent and investment. Like sports and manufacturing, the legal industry may become increasingly technological, statistical, and scientific. This development could make it more difficult for new lawyers to enter the profession, by reducing the number of entry-level jobs.
If we want AI-assisted lawyering to distribute power rather than concentrate it, we need to pay attention to how it is used. We should invest more, not less, in public defenders in order to avoid shackling low-income defendants with more overworked, weakly AI-assisted lawyers. We should discourage the proliferation of AI-initiated lawsuits, liens, evictions, and other legal proceedings. This would require novel solutions; perhaps the penalty for filing frivolous litigation could be raised, or the court filing fee could increase as an entity files more suits. Inevitably, the courts themselves will turn to AI to triage and to automate court proceedings. Courts may even turn to AI to adjudicate disputes.
Excerpted from Rewiring Democracy: How AI Will Transform Our Politics, Government, and Citizenship by Bruce Schneier and Nathan E. Sanders. Reprinted with permission from The MIT Press. Copyright 2025.
Bruce Schneier is an internationally renowned security technologist and the New York Times bestselling author of fourteen books, including Data and Goliath and A Hacker’s Mind. He is a Lecturer at the Harvard Kennedy School, a board member of EFF, and Chief of Security Architecture at Inrupt, Inc. Find him on X (@schneierblog) and his blog (schneier.com).
Nathan E. Sanders is a data scientist focused on making policymaking more participatory. His research spans machine learning, astrophysics, public health, environmental justice, and more. He has served in fellowships and the Massachusetts legislature and the Berkman-Klein Center at Harvard University. You can find his writing on AI and democracy in publications like The New York Times and The Atlantic and at his website, nsanders.me.