Rewiring Democracy

1. AI is Fundamentally Rewiring Democracy as an Information System

Democracy has always been intertwined with technology.

Historical context. From ancient Athens' kleroterion to modern voting booths, technology has always been integral to how democracies function. AI is the latest, most complex information-processing technology poised to reshape how individual preferences are translated into group policy and action. This transformation is not merely incremental; AI's "pluripotent quality" allows it to turn speech into action, akin to money's pervasive role in politics.

Information flows. Democracy, at its core, is an information system. It manages flows of information from citizens to representatives, from data to policy decisions, and from decisions to execution. AI, with its ability to synthesize, describe, predict, and decide, directly infuses these information flows, addressing the "information-processing capacity" gaps that plague modern democracies.

Legitimacy at stake. As AI becomes embedded, governments must decide how much legitimacy to confer upon these tools. Public consent is crucial for democratic legitimacy, and AI systems developed or used in corrupt, unjust, or overreaching ways risk undermining this trust, potentially leading to backlash or even revolution. Authoritarian regimes, unconstrained by public consent, can exploit AI for surveillance and repression, highlighting the divergent paths for AI adoption.

2. AI's Superhuman Capabilities Drive Pervasive Transformation

AI will often not be as effective as a human doing the same job.

Beyond human limits. AI's transformative power stems from its advantages over humans in four key dimensions: speed, scale, scope, and sophistication. These capabilities allow AI to perform tasks that humans are too slow for, execute actions simultaneously across millions of instances, handle a wider variety of tasks than any single person, and consider more factors in decisions than human cognition allows.

Predictive core. At its heart, most AI is predictive, whether forecasting tumor malignancy or generating the next word in a sentence. This predictive power, combined with generative capabilities, enables AI to decide, summarize, and execute tasks autonomously. The utility of AI doesn't require perfection; it merely needs to be "good enough" compared to existing human-mediated systems, which are often imperfect bureaucracies.

Acclimatization and adoption. Public acceptance of AI will grow due to its increasing integration into daily life, demonstrated transparency, and ability to simplify tasks. However, the ultimate driver of AI adoption in governance is often the belief of powerful individuals—legislators, partisans, judges—that it is useful to them, leading to its deployment regardless of broader public trust or ethical concerns.

3. AI Inherits and Amplifies Human Biases and Values

Any AI will have an embedded value system—and biases—because it is created by humans and trained on human-generated and -curated data.

Inherent bias. AI systems are not neutral; they reflect the values and biases of their human creators and the historical data they are trained on. This means AI can exhibit racism, sexism, ableism, and other forms of discrimination, sometimes unintentionally, but also through deliberate "empiricism-washing" to disguise partisan or self-serving outcomes as objective.

Amplified injustice. When deployed in critical areas like hiring, security, or legal sentencing, AI can amplify and entrench patterns of bias and discrimination at unprecedented speed and scale. This raises concerns about "automating inequity," where historical injustices are codified and accelerated by technology, making it harder for marginalized groups to achieve fair treatment.

Controlling values. The ability to easily swap AI systems means a new administration can rapidly implement its definition of fairness or policy priorities, potentially leading to swift ideological shifts in bureaucracy. While this offers efficiency, it also concentrates power at the top, making it easier for leaders to embed their values without the inertia of a human civil service. However, AI also offers new tools for exposing and correcting bias through rigorous, transparent testing.

4. AI Will Revolutionize Political Campaigns and Legislative Processes

AI gives candidates, political advocacy organizations, and even nonpartisan organizations the opportunity to take their communications to the next level.

Personalized politics. AI enables unprecedented personalization of political speech, allowing candidates to tailor messages to individual voters, conduct simulated polling, and manage campaigns at superhuman scale. This can increase voter access to information and help candidates build new coalitions, but also empowers demagogues to spread "custom lies" and manipulate the electorate with targeted misinformation, exacerbating "flooding the zone" tactics.

Legislative efficiency. In lawmaking, AI can assist legislators by drafting bills, identifying loopholes, ensuring consistency, and summarizing complex documents. This can reduce reliance on lobbyists, lower the cost of lawmaking, and help legislatures counterbalance the executive branch by writing more precise laws. However, it also risks creating laws that are too complex for humans to fully understand without AI assistance, and could be exploited for "microlegislation" benefiting special interests.

Coordination and strategy. AI will become a crucial tool for campaign coordination, volunteer management, and strategic negotiation. It can help political movements organize across diverse jurisdictions and languages, but also facilitate "astroturfing" and foreign influence operations. The ability of AI to analyze vast amounts of data and simulate outcomes will make political strategy more sophisticated, potentially benefiting those with the best AI tools and data.

5. AI Will Reshape Government Administration and Legal Systems

Automating judicial tasks throws these profoundly human decisions into the control of machines.

Streamlining public services. AI offers significant potential to enhance public services by speeding up processes like language translation, traffic optimization, benefits administration, disaster management, and FOIA requests. For example, AI can drastically reduce the time needed to process disability claims or review property records, making government more efficient and accessible. However, poor implementation or inherent biases can lead to "automating inequity," as seen in faulty welfare algorithms.

Operating and enforcing. Governments are rapidly adopting AI for internal operations, including digitizing archives, auditing for fraud, managing logistics, and collecting taxes. AI can make regulatory enforcement more uniform and thorough, potentially leveling the economic playing field by ensuring all firms play by the same rules. Yet, this also raises concerns about totalitarian surveillance, the potential for AI to be hacked or manipulated, and the risk of automating politically motivated selective enforcement.

Judicial transformation. In the legal system, AI is already augmenting lawyers for research, drafting, and jury selection, and assisting judges in summarizing briefs and composing rulings. AI arbitrators could offer speedy resolution for quotidian disputes, but raise questions about bias and oversight. The ability of AI to interpret legislative intent and even encode laws in computer language could shift power between legislative and judicial branches, but also risks creating incomprehensible laws and eroding human deliberation.

6. AI Empowers Citizens, Yet Risks Concentrating Power

AI gives people—and corporations, and other organizations—the capacity to control a nearly unlimited number of agents.

Power amplification. AI acts as a powerful amplifier, endowing individuals and organizations with superhuman capabilities across speed, scale, scope, and sophistication. This means that while AI can broadly empower citizens by making government more accessible and responsive, it often magnifies existing power imbalances, benefiting the wealthy and powerful disproportionately.

Citizen advocacy. AI can help citizens advocate more effectively by transforming policy preferences into articulate communications, submitting testimony, and navigating complex bureaucracies. Tools like Resistbot demonstrate how AI can lower barriers to political participation, especially for people with disabilities or limited time. However, this also changes the "signaling function" of political engagement, as mass-produced messages may dilute the impact of genuine, deeply felt concerns.

Personal political proxies. In the long term, AI could act as personal political proxies, continuously discerning and advocating for individual preferences, potentially enabling new forms of direct democracy with hundreds of ballot initiatives. This could make government more responsive and equitable by overcoming limitations of traditional representation. Yet, it risks the "atrophy of our innate capability for democracy," as human deliberation and personal conviction could diminish if AIs do all the political heavy lifting.

7. Trustworthy AI is Essential for Democratic Governance

Trust is essential to a functioning society.

Foundational requirement. For AI to benefit democracy, it must be trustworthy. This encompasses accuracy, reliability, fairness, security, and privacy. Without these, citizens will lose faith in government functions and policies, leading to demands for reform or even regime change. The scope of trust required varies by application, from a candidate trusting their chatbot to society trusting an AI benefits administrator.

Barriers to trust. Numerous factors impede trust in AI:

• Mistakes: AI's "weird and nonintuitive" errors, hallucinations, and confident delivery of incorrect information.
• Bias: Inherent biases from training data and human creators, leading to unacceptable discrimination.
• Insecurity: Vulnerability to hacking, manipulation, data leaks, and disruption, undermining data integrity.
• Untrustworthy developers: Corporate profit motives can lead to AI systems that prioritize monetization over user interests, potentially manipulating users or reflecting corporate biases.

Beyond technology. The challenges to trustworthy AI are more social and political than technical. While safeguards like transparency, disclosure, and robust appeals processes can help, the fundamental issue is who controls AI development and deployment. The current concentration of AI control among a few powerful corporations poses a significant threat to democratic values.

8. Democracy Must Actively Steer AI for Public Benefit

To protect the interests of all members of a pluralistic society, democracies must look beyond the isolated, technical aspects of AI to recognize that social policy is technology policy.

Beyond regulation. Governments must actively steer AI development and use towards ethical, just, and equitable outcomes, not merely regulate its technical minutiae. This requires a holistic approach where "social policy is technology policy," recognizing that AI impacts existing social behaviors and power structures. Regulations should focus on specific activities and their outcomes, rather than rapidly evolving AI capabilities.

Public AI alternatives. A critical step is to diversify control of AI beyond private, for-profit corporations. "Public AI," developed under democratic control and for public benefit, offers a trustworthy alternative. Governments, from national to local levels, can leverage open-source components to build AI models aligned with public interest, setting a competitive baseline for transparency and responsiveness that commercial entities must meet.

Resistance and renovation. Citizens and institutions must actively resist harmful uses of AI, such as unethical applications or totalitarian surveillance. Simultaneously, democracies need "renovation" – adopting reforms that address existing flaws exacerbated by AI, like strengthening civil service protections, enhancing court capacity, and limiting lobbying. The Writers Guild of America strike serves as a blueprint for how to build power and secure protections in an AI-transformed economy.

9. Principles for Democratic AI Guide Responsible Development

To build AI that works for democracy and that can bring about more benefit than harm, we need organizing principles under which each party involved in developing and operating AI can operate.

Shared vision. A shared vision for democratic AI is crucial, guided by principles that ensure AI serves the public interest. These principles are:

• Broadly capable: AI must adequately perform a full range of tasks, from prediction to generation.
• Widely available: Accessible to all entities regardless of identity, affiliation, or wealth, with unrestrictive licensing and low cost.
• Transparent: Open to review, testing, and extension, with open-source code and data.
• Meaningfully responsive: Developers must actively solicit and demonstrably respond to stakeholder input, monitoring and mitigating disparate impacts.
• Actively debiased: Human choices drive AI biases, which must be acknowledged, managed, and corrected, without denying users' values.
• Reasonably secure: Reliable, non-exploitative, protecting confidential information, enforcing data integrity, and uncorrupted by private interests.
• Nonexploitive: Must not co-opt public labor or resources for private profit, internalizing environmental and social costs.

Overcoming barriers. While broad capability exists, corporate AI models often incorporate structural incentives that limit availability, transparency, responsiveness, and non-exploitation. Overcoming these barriers requires governments to rigorously regulate AI developers, and for the market to incentivize trust as a valuable asset. The goal is not perfect AI, but AI that is demonstrably better and fairer than existing human systems, and that empowers all citizens.

Last updated: March 31, 2026