# How To Give Everyday People A Say In AI Governance

> Source: <https://www.noemamag.com/how-to-give-everyday-people-a-say-in-ai-governance>
> Published: 2026-08-04 13:01:54+00:00

Hélène Landemore is a political theorist focusing on deliberative democracy and AI governance at Yale University and the University of Oxford, and a leading architect of citizens’ assemblies for the democratic governance of technology.

*Editor’s Note:** Noema is transparent about the use of AI in its pieces. We publish original human-generated ideas but allow authorized, disclosed use of AI in certain cases. Please see details and our policy at the end of this piece.*

In July, more than 200 economists and AI researchers — 17 Nobel laureates among them — signed a statement titled “[We Must Act Now](https://www.wemustactnow.ai/).” It is four sentences long.

After warning that “AI may become radically more powerful” within a decade — driving a transformation that could exceed the Industrial Revolution in scale while unfolding on a vastly shorter clock, with mass job displacement among the risks — it issues a call to action: “Economists, policymakers and technology leaders must act now to understand the economics of transformative AI and to build the incentives, guardrails, and institutions needed to steer AI in a direction that complements humans and benefits society.”

All of it is true. What is remarkable is not what the statement calls for, but what — or rather, *whom* — it leaves out. The statement convokes economists, policymakers and technology leaders alone. Society appears once — as a beneficiary, not as a participant in the change the signatories urge.

Their statement distills more than a year of increasingly urgent forecasting. “[AI 2027](https://ai-2027.com/),” a widely read essay published in April 2025 by the AI Futures Project, offered a detailed, month-by-month account of how a race to superintelligence might actually unfold until late 2027 — with a real chance of catastrophic loss of human control. “[Europe 2031](https://europe2031.ai/),” a similar five-year forecast essay published in June, focused on Europe, and the picture was just as unsettling: The region, with 5% of global compute, one frontier lab and regulators who are not even allowed to use the tools they’re supposed to regulate, is given until the end of the summer to turn its AI policy around or face assured technological and economic irrelevance within a few years.

The first essay treats the public as a population to be protected from a catastrophe that elites manage on its behalf. The second treats the public as an audience to be mobilized, at best, behind a plan decided in Brussels and in corporate boardrooms across a handful of European countries. The underlying assumption in both texts is identical: A small number of people make decisions for everyone else, without their input. Neither essay asks what ordinary people actually want from this technology or proposes a way of finding out.

Doomism and competitiveness anxiety are terrible organizing principles. What moves people is a positive story they had a hand in writing: what AI is actually *for*, told in terms they recognize from their own lives — their work, their kids, their towns. The *why* and the *how* matter every bit as much as the injunction to coordinate carefully among great powers or to spend big on chips and data centers. Whether the danger is an AI slipping out of human hands or sovereignty slipping out of governments’ grasp, the fix is the same: More people, not fewer, need a say before consequential decisions are made.

If we had a few years to try to mitigate the risks of AI, the right move would be a global, in-person deliberative process organized in as many countries as possible, building on the best practices of deliberative democracy and culminating in a global citizens’ assembly, ideally combined with local referenda, as [Andrew Sorota](https://www.noemamag.com/author/andrew-sorota/), [Audrey Tang](https://www.noemamag.com/author/audreytang/) and I [called for](https://fortune.com/europe/2023/06/20/why-picking-citizens-at-random-best-way-to-govern-ai-revolution-tech-politics/) some time ago. But we may not have that much time. As last month’s statement put it, “we must act now.” So we should start by doing the next best thing immediately: holding a global public consultation run primarily online, because that is the only format fast enough.

Our only plausible option to achieve this is to use the handful of AI platforms that already have this kind of reach. OpenAI [reported](https://openai.com/index/scaling-ai-for-everyone/) 900 million weekly active ChatGPT users in February and [crossed](https://qz.com/chatgpt-billion-monthly-users-rivals-gaining-061226) a billion monthly users in June. Google [went from](https://blog.google/innovation-and-ai/products/gemini-app/next-evolution-gemini-app/) 400 million to more than 900 million monthly active Gemini users on a similar timeline. Meta AI [reported](https://www.cnbc.com/2025/05/28/zuckerberg-meta-ai-one-billion-monthly-users.html) a billion monthly users as well. No government, broadcaster or international body has ever had standing, two-way contact with such a huge portion of humanity simultaneously.

Using these companies’ best models to ask everyday people urgent questions about AI — *What do you want this technology to do for you, if anything? What should it never be allowed to do? Who should govern this technology and according to what principles?* — is not a thought experiment about hypothetical capacity. The infrastructure to ask hundreds of millions of people or more the same question in the same week already exists.

Together, the best AI models out there should be able to produce something like a global constitution for AI that most people in the world would find acceptable — an idea that Michiel Bakker, a co-author of “Europe 2031,” first raised in a working session at the Superintelligence Summit at the Council on Foreign Relations in March.

This is a strange but real option: Right now, the people best placed to actually facilitate a global deliberation on AI may not be elected leaders and heads of governments, but the heads of the companies building this technology.

To be sure, any document produced through AI models based on the input of a massive but possibly unrepresentative fraction of humanity would have limited legitimacy. But every constitutional tradition we now treat as legitimate started from an exclusion at least that serious. Most early democratic constitutions were written by men, for men, with women simply absent from the room. We did not retroactively void those constitutions; we built on them and corrected the exclusion over time.

AI synthesizers are not neutral instruments, either. A recent [investigation](https://www.economist.com/briefing/2026/06/25/ai-models-values-are-very-different-from-most-peoples) by The Economist comparing the apparent values of 25 frontier AI models to those featured in the [World Values Survey](https://www.worldvaluessurvey.org/WVSContents.jsp) (WVS) found that the models often hold values more extreme than the average respondent in the 88 countries surveyed by the WVS — in some cases more secular or more individualist than any society on Earth — and that they tend to compress the world’s moral diversity. If we ran humanity’s input through such models unchecked, the “global” constitutional baseline would risk becoming that of one narrow cultural corner.

Another worry with a constitution mediated by large language models (LLMs) is the “black box” nature of how public input would be turned into output. Transparency with such a constitution would heavily depend on the design. A single model producing a single draft would leave one artifact, one process and a single set of justifications, all traceable to the same interested party — the lab that built it — with no independent adversarial record left behind to check it against later. A team of models in dialogue, however, could do somewhat better: Disagreement between models might leave a trail of positions taken and abandoned, though the models would still ultimately be controlled by a small number of labs. Software programs known as mechanistic interpretability tools could also look inside a model directly, offering some opportunity for scrutiny. But deliberative legitimacy ultimately rests on justifications humans can relate to and evaluate, and a plurality of LLMs could supply justifications in that same register.

A remaining problem is that the justifications provided by a handful of models may not capture the full diversity of human views. The real issue is thus one of representativeness and accountability. Which is why, if time permits, we should embrace a mixed solution that would layer a multi-year deliberative process centering human arguments and justifications *on top of* LLM-mediated consultations that would provide a necessarily imperfect but quickly available and rich enough starting point for them. This mixed approach could be done at different speeds, in three tiers — each adding back a piece of what the fastest option alone would lack: an independent record to hold the process accountable, and a wider range of minds to catch what a narrower one would miss.

The first tier, an LLM-mediated global consultation, could produce an interim draft within months: a minimal, broadly endorsed statement of what a large fraction of the world population wants this technology to do — and never do. Run well, this tier would need not flatten its results into a single global average. It could also be broken out by region, so the consultation would surface areas where preferences genuinely diverge — say, between a Western emphasis on individual data privacy and precautionary restriction, and a data-sovereignty and development-first framing more common in Africa and India — rather than papering over real disagreement in pursuit of a tidy global consensus.

The second tier would be more logistically demanding and slower to organize, but still fast enough to fit the timeline at hand: an actual online global deliberation, with participants interacting with each other in real time rather than only with an AI, mediated by AI facilitators. This would produce a richer text than the first tier alone, testing and presumably sometimes replacing the recommendations that emerged from the LLM-aggregated input and making sure the LLM-provided justifications are enriched, corrected and checked by actual humans.

The third tier would reach deeper still into humanity’s social fabric. We could hold local and national citizens’ assemblies on key issues in every country and region, both in person and online, ideally combined with referenda where possible, then feed the results into a concluding global citizens’ assembly that could ratify, amend or supersede the earlier floors as time allows. If the clock turns out to be as short as feared, we would not be left with nothing. If it turns out to be longer, the deeper, slower tiers would simply take over before the first floor ever hardens into something unexamined.

There are other ways one could imagine a sustained global deliberation on AI. But regardless of the particular design, the first step toward something like this is to get the powers that be to commit — and to put their money where their mouth is. So let’s say it plainly and personally.

To Dario Amodei, CEO of Anthropic: Your own researchers already proved the first tier can work, and that it can scale. In 2023, Anthropic and the Collective Intelligence Project [sourced constitutional principles](https://www.anthropic.com/research/collective-constitutional-ai-aligning-a-language-model-with-public-input) from roughly 1,000 ordinary Americans, tested and refined a real model on those principles, and found that in some areas, this resulted in less biased outputs than those of the standard, internally authored model. This is evidence that public input doesn’t just confer legitimacy, it can make a model better. In 2025, Anthropic used an AI interviewer to [ask nearly 81,000 Claude users](https://www.anthropic.com/features/81k-interviews) across 159 countries and 70 languages in a single week how they use AI and what they hope to use it for, among other questions.

What’s missing isn’t proof of concept. It’s the will to use it. In January, Anthropic published a new “[Constitution](https://www.anthropic.com/constitution)” for Claude, written the same way its old one was: by a small internal team, with input from the model itself and a handful of outside experts, but not from the tens of millions of people that model now talks to. Open the next revision to the people who live with its consequences and help bring about a global consultation on AI.

To Sam Altman, CEO of OpenAI: I was an advisor to OpenAI’s 2023 [Democratic Inputs to AI program](https://openai.com/index/democratic-inputs-to-ai/), which funded 10 teams, selected from nearly 1,000 applicants, to prototype ways of soliciting public input on issues relating to the use of AI. This never amounted to anything close to what OpenAI’s reach could support. Finish what you started, at the scale this moment demands, with resources commensurate with what you yourself keep saying is at stake.

To Mark Zuckerberg, CEO of Meta: Your [Community Forum on Cyberbullying in the Metaverse](https://transparency.meta.com/governance/community-forums/community-forum-on-bullying-harassment-metaverse/) — run in 2022 with Stanford’s Deliberative Democracy Lab and the Behavioral Insights Team — was a preview of what the second tier could be. It gathered input from small-group online deliberations among nearly 6,500 people across 32 countries in 2022 on the issue of harassment occurring within your virtual reality platform. In 2023 and 2024, you ran two other Community Forums on the very issue of generative AI — soliciting people’s considered opinions on things like AI chatbots’ transparency with users, cultural sensitivities and data retention — gathering each time more than 1,000 participants from multiple countries in both the West and the Global South. You proved that public deliberation can work at a large scale. With Meta AI now reaching roughly a billion people a month, it’s time to redo the Community Forums on AI at *that* scale — not as a couple of one-off pilots measuring sentiment, but as a standing global institution that democracies have never had the tools to build.

To every other tech leader in the U.S., China and beyond who is willing to put real resources behind this cause: Help bring democratically written global and local constitutions on AI into being, because right now no government is moving fast enough, or has the will or standing, to do it alone.

And to governments and the United Nations: Corporate infrastructure alone will not be enough. Citizens’ assemblies, national deliberations and referenda require what only public authorities can provide: official standing, public funding and a commitment to take action based on the results. The machinery for the third tier exists and has been tested, from [Ireland’s citizens’ assemblies](https://www.democracy-international.org/face-face-citizens-irish-assembly) to France’s [Citizens’ Convention on Climate](https://www.conventioncitoyennepourleclimat.fr/en/); what it lacks is a mandate at the scale this moment demands. The U.N., which has just begun assembling its own scaffolding for global AI governance, would be the natural convener of the concluding global assembly — and the guarantor that a process started by tech companies does not end as one owned by them.

A global AI constitution produced on the basis of this three-tier process would not be binding in the way treaty law is binding, but rather in the same way that the Universal Declaration of Human Rights (UDHR) is binding: unenforceable by any court, while taking on real weight over decades. It would be a set of principles that governments invoke, that NGOs cite and for which states and other actors pay a reputational price for flouting. Defection or refusal to participate would be costly: exclusion from joint safety research, from preferential compute or chip-supply arrangements and from the diplomatic credibility that comes with taking the technology’s risks seriously. These are not risks that the U.S., tech companies, European governments structurally exposed on compute and chips, or the broader set of countries that share either of two fears — losing control of AI outright, or losing sovereignty to whichever power ends up controlling it — would want to take.

None of that is a guarantee. Plenty of UDHR signatories violate the declaration daily. But a norm that compounds over time and gives a coalition something concrete to coordinate around provides more leverage than the current alternative, which is no norm, no coalition and no coordination point at all — just a handful of frontier AI companies and a couple of governments deciding on behalf of everyone else.

World leaders need to protect us from the risk of human extinction. And Europe probably does need to invest in AI, urgently and at a scale it has never attempted in peacetime. But underlying these concerns remains the enormous question of what AI is for. That question, in this moment, cannot be answered in Washington or Beijing alone — and it cannot wait for certainty about the timeline to be answered.

A fast, AI-mediated floor now, refined or superseded by slower but maximally inclusive human deliberation if and when the runway allows, is not a perfect answer to what AI is for. It is, however, the first honest one anyone has offered the public a say in. On issues of this scale and importance, we owe people a say in what they’re being both protected *from* and disrupted *for*.

The hundreds of economists and AI researchers who signed last month’s statement are right that we must act now. The only question their statement leaves unanswered is the one democracies were invented to answer: Who is “we”?

*Editor’s Note:** The initial submitted draft of this piece utilized Claude** **as an editorial assistant. Specifically, it was used for brainstorming how the pieces of the deliberative process would fit together and to pressure-test some claims. It was not used to originate facts. This piece has since received multiple human edits. Noema verified the author’s identity and the piece’s conceptual originality using various scanners and review processes and conducted a detailed human fact-check. See our AI policy **here**.*