Who Agreed to This? Consent, Democracy, and the Governance of AI

When a technology reshapes work, information, and daily life within a few years, a fair question follows: who decided it should happen this way, and were ordinary people ever asked? The honest answer is that AI governance is being built in real time, across many venues at once, and the mechanisms for public consent are still thin and contested. What follows is a map of who is making the rules, and where the public voice enters — or doesn't.

The rules being written now

The European Union has moved furthest toward binding law. The EU AI Act entered into force on 1 August 2024 and phases in over several years: bans on certain 'unacceptable-risk' uses and AI-literacy duties began in February 2025, obligations for general-purpose AI models began in August 2025, and most high-risk rules and enforcement begin in August 2026. Penalties can reach €35 million or 7% of global turnover.

The United States has taken a markedly different path. In January 2025 the administration issued an executive order titled 'Removing Barriers to American Leadership in Artificial Intelligence,' revoking the prior administration's 2023 AI order, and in July 2025 published 'America's AI Action Plan,' written to remove perceived regulatory obstacles. A December 2025 order sought to limit state-level AI regulation. A separate proposed ten-year moratorium on new state AI laws passed the House but was rejected by the Senate amid bipartisan concern over state authority. The result is a live contest between a federal 'light-touch' preference and dozens of individual state laws.

The international layer

Alongside national rules, a series of summits began at the UK's Bletchley Park in November 2023, continued in Seoul in May 2024, and became the Paris 'AI Action Summit' in February 2025. Observers noted the framing shifted over that arc from 'safety' toward 'action' and investment; the Paris meeting drew delegates from over 100 countries and produced a declaration on inclusive and sustainable AI, though the US and UK declined to sign it.

At the United Nations, the September 2024 Global Digital Compact committed members to new machinery, and in August 2025 the General Assembly adopted resolution A/RES/79/325 by consensus, establishing an Independent International Scientific Panel on AI and a Global Dialogue on AI Governance, with the first Dialogue session set for Geneva in 2026. These bodies advise and convene; they do not make binding law.

Where does the public come in?

This is the crux of the 'democratic deficit' debate. Most governance so far runs through executives, regulators, legislatures, and expert panels — the ordinary channels of representative government and diplomacy. Direct public input has been comparatively limited.

Survey evidence suggests a gap between public sentiment and the direction of policy in some countries. A 2025 survey by the Ada Lovelace and Alan Turing Institutes found 72% of the UK public said laws and regulation would increase their comfort with AI, up from 62%, and nearly nine in ten wanted AI to be independently regulated. A global Carleton–CIGI–Ipsos survey similarly reported strong majorities favoring coordinated regulation over prioritizing rapid innovation.

Some efforts aim to add a deliberative public voice directly. Citizens' assemblies — randomly selected panels that study an issue and issue recommendations — have been proposed as a way to give informed public input on AI, and coalitions have run deliberative dialogues around recent summits. These remain advisory and small in scale relative to the deployment they respond to.

Two honest positions

The disagreement is genuine. One view holds that heavy or premature rules risk slowing a fast-moving field, entrenching incumbents, and ceding advantage to less cautious competitors; on this account, markets, iteration, and targeted after-the-fact rules serve the public better than up-front restriction. The opposing view holds that a technology this consequential should not be deployed at population scale without stronger, earlier democratic input, and points to survey data showing public appetite for guardrails.

Both sides can claim principle. The unsettled questions are concrete: Should binding rules precede deployment or follow harm? Which decisions belong to legislatures, which to regulators, and which — if any — to citizens directly? And who counts as 'the public' for a technology whose effects cross every border? Those questions are being answered right now, mostly by institutions, and the record of whether ordinary people were meaningfully consulted is still being written.