I recently read the The Intelligence Curse, by Luke Drago and Rudolf Laine. I found the series of essays very interesting and at the same time questioning. They sound an alarm that also sometimes make me feel ill at ease: once Artificial General Intelligence becomes “general intelligence on tap,” capital can copy super-human labour instead of hiring people, and “powerful actors … will lose their incentives to invest in people.” The authors liken that dynamic to the resource curse, where oil revenues let rulers ignore citizens and still stay in power; the data show such states suffer “poorer economic growth and higher rates of poverty.”

Drago & Laine’s remedy has three pillars—Avert catastrophic AI, Diffuse capability so individuals remain productive, and Democratise institutions. I love those three call to actions, but the pages on democratization are – in my view – missing a full world of governance innovations that have been surging since two decades. They consider that existing decision mechanisms like representative structures or voting could adapt to resist the Intelligence curse.

I argue in this essay that they don’t and that we already have a better way: sortition-based governance and standing Citizens’ Assemblies, which offer a faster, fairer, and more tamper-proof way to keep humans inside the loop of power and curse the curse.

Election calendars crawl while transformer checkpoints leap. Legislatures are overloaded with lobbying and partisanship and take years to digest what AI labs unveil in weeks. The result is a widening accountability gap at precisely the moment power is concentrating in owners of compute.

Real-world mini-publics already outperform slow representative governments on complex issues:

• France’s Citizens’ Convention for Climate (2019-20) was hailed as “a groundbreaking political experiment that placed citizens’ voices at the centre of national conversations on climate policy.”(publicdeliberation.net) Its 150 jurors met for seven long weekends and produced 149 legal proposals, more than 60% of which are now implemented in policy.

• European Citizens’ Panels organized by the three EU institutions and later by the European Commission already assembled around 2000 randomly selected citizens from across the EU and proved again and again (10 such panels have been concluded) that citizens could successfully deliberate at a pan-European level.

Random selection (sortition) delivers diversity, shields against capture, and—because each panel can reconvene at very quick pace —matches the tempo of technological change.

Sortition scales beyond a single assembly. Parallel juries can each focus on a slice of the AGI problem-space—Safety Review, Economic Transition, Rights & Justice, Global Commons—with a meta-jury to resolve conflicts and an open ledger that records every hearing, vote, and expert brief.

Drago & Laine insist on “diffus[ing] AI-enabled technology that augments human productivity and keeps humans in the loop.” In practice, assembly members can use open-source language-model assistants to translate expert evidence instantly, surface hidden counter-arguments, and run scenario simulations—turning AI from job-killer into deliberative amplifier while remaining auditable by the citizens themselves.

If the intelligence curse is a future where wealth, knowledge, and political clout pool inside sealed servers, the antidote is a governance stack that is open, distributed, and relentlessly upgraded by the people it serves. Sortition-based Citizens’ Assemblies—augmented by AI yet sovereign over it—offer that stack. The practice exist, the theory is sound, and the need is urgent.