Rewiring Democracy: How AI Will Transform Our Politics, Government, and Citizenship
Robert Diab
This book makes an incisive and technically informed contribution to a rapidly growing body of work on AI and politics. Its central chapters survey the tangible ways AI will alter democratic practices, including political campaigning, lawmaking, administration, and legal judgment. The book is framed by the authors’ conception of democracy as “an information system that translates individual preferences into group policy decisions” (p.27), a system that can be optimized through effective use of AI. This conception has its merits, but it underplays democracy as a normative order structured by constitutional limits, institutional friction, public reasoning, and commitments to political equality—and leads the authors, in places, to understate AI’s potential impact on some of its most important dimensions.
This detailed analysis of AI’s political ramifications can be placed alongside recent work that includes Mark Coeckelbergh’s Why AI Undermines Democracy and What To Do About It (2024), an edited collection AI and the Future of Democracy (2026), and a Carnegie Endowment report, AI and Democracy: Mapping the Intersections (2026).1 Schneier and Sanders approach the topic not as political theorists but as technologists with “one foot in the academy and the other in industry” (p.5). Schneier, a leading cybersecurity thinker, and Sanders, a data scientist, bring an unusually strong grasp of AI’s functional strengths and weaknesses to the topic and of how these may be harnessed or abused in the practice of politics. Their treatment of AI’s impact on democracy is similar in approach but more extensive than that found in the Carnegie report. Schneier and Sanders draw less on political theory and history than Coeckelbergh does in his book. And while the authors are alive to ethical dimensions of using AI, this features less prominently in Rewiring Democracy than it does in the edited collection noted here.
Three assumptions frame Schneier and Sanders’ analysis. The first is that technology has always played a role in shaping democracy. From the kleroterion of classical Athens (an early lottery apparatus for allocating public office), to the railway, radio, television, and the internet, technologies have long mediated political participation and authority. AI, they argue, belongs in this lineage and is best understood in terms of continuity rather than rupture. But the analogy to earlier media risks overlooking what is genuinely new about AI and what may be potentially more threatening to democracy. Language models transmit information more efficiently but also act as interlocutors, agents, and more sophisticated intermediaries, with arguably greater potential to distort democratic communication than earlier media.
Their second assumption is that technological change tends to come from the bottom up rather than through large state projects. AI, in their view, will diffuse incrementally into institutions and everyday practices, much as the internet did. Yet the claim sits uneasily with the present political economy of AI, where the most powerful systems are concentrated in a small number of firms whose control over infrastructure increasingly rivals that of states.
The third and most consequential assumption is that democracy is, as noted, an information system for turning individual preferences into policy decisions. AI matters because its strengths—the authors single out speed, scale, scope, and sophistication—vastly enhance the ability to process information. Conceiving of democracy in this way is helpful to some degree; it places a focus on how AI might improve or impede institutions in specific ways. But it also narrows the field of vision in that democratic institutions do more than process information efficiently. They also provide a forum for productive disagreement that leads to compromise, and a space in which citizens interact with one another as political equals. Friction and even certain kinds of inefficiency, such as a prolonged public debate or contentious constitutional litigation, are often signs of a thriving democracy.
These tensions become clear in the book’s most substantive and illuminating chapters, which examine how AI may reshape specific sites of governance. In their discussion of campaigning, Schneier and Sanders argue that AI will transform electoral politics through automated translation, hyper-personalized messaging and conversational systems capable of engaging voters at scale. These tools lower barriers to participation for a wider range of actors and could help candidates reach constituencies previously beyond their reach. The authors are alert to the dangers of these tools, including generating propaganda as easily as persuasion, and personalized messaging further fragmenting the public sphere. Yet they remain optimistic that AI-assisted conversation between candidates and voters could deepen democratic engagement. While it might do so in some respects, it would also likely accelerate the fragmentation well under way.
Schneier and Sanders’ treatment of lawmaking raises similar concerns. They imagine legislators using AI to help sort and consolidate diverse voter preferences and goals on a greater scale than was possible before. Lawmakers could also use AI to draft long and complex statutes with greater clarity and internal consistency. On reflection, these use cases seem doubtful. Legislation on complex or contentious matters involves more than issues of information management. Lawmakers reach compromises through negotiations that are often messy and chaotic—and, in common law jurisdictions at least, it is not uncommon to leave key provisions vague or ambiguous, for courts to sort out later. The complexity and compromise inherent in democratic lawmaking, and its often imperfect results, cannot be engineered away; AI’s role here may turn out to be more limited than the authors suggest.
The chapter on administration offers a more plausible assessment, partly by resisting easy comparisons between flawed AI and idealized human decision-making. Schneier and Sanders note potential risks and misuses of AI in this context, invoking the notorious Australian “Robodebt” episode, in which hundreds of thousands of citizens were mistakenly issued benefit repayment notices. Yet the relevant question, in their view, is not how AI compares to an ideal or perfect process but how it compares to existing bureaucratic processes, which are they frequently opaque, biased or error-prone. Government agencies could use AI effectively, they argue, if guided by frameworks that foster transparency and accountability, and if they build in robust appeal mechanisms. A counterargument here the authors do not fully address is whether confidence in the legitimacy of administrative justice can be sustained when decisions are made on a probabilistic basis by systems whose underlying processes remain opaque, even where procedural protections are in place.
The chapter on AI’s potential role in assisting courts carries over an optimism that understates the possible impact on perceptions of legitimacy. Schneier and Sanders argue that AI used in legal judgment might be faulted for being opaque, but so too, in many cases, is human judgment. Following an argument Eugene Volokh made in 2019,2 the authors contend that what citizens want is not direct access to mental processes underlying a decision but cogent and consistent justifications for outcomes grounded in valid legal criteria—something language models can now provide. On this view, courts and mediators could use AI in a manner compatible with democratic values if its role is disclosed, its criteria of judgment are transparent, the parties consent to its use, and robust avenues of human review are put in place.
However, the opacity of human judgment is different from that of AI. As critics of automated judgment have argued, we perceive a judge’s written decision or a jury’s verdict as legitimate partly because it reflects the judgment of human decision-makers who render judgment not through a statistical or predictive algorithm but through a holistic, often morally and socially informed assessment.3 For this reason, it is not clear whether the safeguards that Schneier and Sanders point to (disclosure of AI use, transparent criteria, and meaningful avenues for appeal to a human decision-maker) would suffice to maintain confidence in a court system in which AI were to play a substantial role.
In the final chapters, Schneier and Sanders turn to mitigation and reform. We should regulate the institutions deploying AI rather than only model architecture, reduce dependence on private firms by building more public systems, resist abusive uses such as facial recognition in policing, and consider broader responses to the concentration of economic power in the AI sector, including a ‘tech impact tax.’ These are sensible recommendations, and they reinforce the book’s welcome refusal of technological determinism. Nothing about AI is predetermined; we can still shape how it will be used.
The authors’ guarded optimism is one of the book’s main virtues. Rewiring Democracy may understate AI’s impact on some of the normative dimensions of democratic life, but its close attention to how AI is likely to be deployed illuminates how much is already changing. One comes away with a clearer sense of how AI intersects with democratic institutions and how potential misuses might be addressed through reform. The book also raises important questions about whether the informational model of democracy fully captures what is at stake as AI becomes more deeply embedded in our politics.
Footnotes
1. Mark Coeckelbergh, Why AI Undermines Democracy and What To Do About It (Polity Press 2024); J Udo-Udo Jacob and NR Vajjhala (eds), AI and the Future of Democracy: Building Resilient and Inclusive Societies (Routledge 2026); and R George and I Klaus, AI and Democracy: Mapping the Intersections (Carnegie Endowment for International Peace 2026).
2. Eugene Volokh, “Chief Justice Robots” (2019) 68 Duke Law Journal 1135.
3. James Grimmelmann, Blake LW Sobel and Daniel Stein, “Generative Misinterpretation” (University of Wisconsin Legal Studies Research Paper No 1878, 2025) <https://ssrn.com/abstract=5309575> accessed 18 April 2026; and AE Afrouzi, “Robots, Thurgood Martian, and the Syntax Monster: A New Argument against AI Judges” (2024) 37 Canadian Journal of Law & Jurisprudence 369.
Categories: Book Reviews, Rewiring Democracy, Text