Skip to content
Citizen Infrastructure

Four futures for AI and democracy

In 2026 the European Commission’s in-house foresight team — the EU Policy Lab at the Joint Research Centre — published a study asking what democracy could look like by 2050 as AI reshapes it. It is deliberately foresight, not forecasting: rather than extrapolate today’s trends, it hunts for weak counter-signals and pushes them to extremes to expose the tensions ahead. Its headline is that AI’s effect on democracy is not technologically predetermined — the same capabilities could deepen participation and transparency or concentrate power and reduce citizens to data points, depending entirely on design and institutional choices. To make those choices visible, it draws four deliberately unranked scenarios.

The four scenarios

Section titled “The four scenarios”
  • AI-Facilitated Direct Democracy — citizens weigh in constantly, even via wearables; AI aggregates their input and can be delegated to; in-person deliberation persists alongside; and elected politicians shift toward facilitator roles. The “full potential” path, though the report is clear that not every part of it is benign.
  • Citizen Pushback — opaque big-tech AI erodes trust until citizens disengage and then rebel, retreating into off-grid, analogue, local self-government. Democracy survives, but fragments unevenly across regions.
  • Technocratic Overreach — “efficiency” and data-driven expertise become the source of legitimacy, hollowing out elected deliberation until the checks and balances quietly dissolve.
  • Big Tech Power Grab — a corporate “observation state” runs a token-based pseudo-democracy tied to a social-and-consumption score, and public institutions — the EU included — become irrelevant.

The tensions underneath

Section titled “The tensions underneath”

The scenarios are built to surface recurring tensions rather than to pick a winner: efficiency versus meaningful citizen agency, the concentration of power, the use of personal data in governance, trust in the information environment, and — the one the report keeps returning to — whether participation is consequential rather than theatrical.

What it points to

Section titled “What it points to”

The study prescribes no single fix; the tensions are offered as “analytical lenses rather than problems to be solved in isolation.” But its concrete suggestions cluster in the darker scenarios: preserve independent, pluralist, citizen-accountable institutions outside the executive; structurally separate the companies that operate digital platforms from those that sell on them; treat essential digital services as critical infrastructure; and classify data by sensitivity rather than leaving it ungoverned. The through-line is that the future here is a choice, and the levers are institutional. It sits alongside the sharper critique in does AI weaken democratic institutions? and the constructive answers of public AI and the 6-Pack of Care.

Sources

Section titled “Sources”